Content uploaded by Peter H Thrall
Author content
All content in this area was uploaded by Peter H Thrall on Jun 11, 2015
Content may be subject to copyright.
1522
American Journal of Botany 90(10): 1522–1531. 2003.
H
ERBARIUM STUDIES ON THE DISTRIBUTION OF
ANTHER
-
SMUT FUNGUS
(MICROBOTRYUM VIOLACEUM)
AND
SILENE
SPECIES
(C
ARYOPHYLLACEAE
)
IN THE EASTERN
U
NITED
S
TATES
1
J
ANIS
A
NTONOVICS
,
2,4
M
ICHAEL
E. H
OOD
,
2
P
ETER
H. T
HRALL
,
3
J
OSEPH
Y. A
BRAMS
,
2
AND
G. M
ICHAEL
D
UTHIE
2
2
Department of Biology, University of Virginia, Charlottesville, Virginia 22904 USA; and
3
Division of Plant Industry, CSIRO, Canberra, Australia
We used herbarium specimens of Silene virginica,S. caroliniana,S. rotundifolia, and S. latifolia to survey the incidence of anther-
smut disease (caused by Microbotryum violaceum sensu lato) in the eastern USA. We found no evidence of a collector bias against
diseased specimens. Diseased specimens were frequently found in collections of S. virginica and S. caroliniana, but not in those of
S. rotundifolia or S. latifolia. Disease incidence in S. virginica and S. caroliniana increased significantly over the past century and
was higher in marginal populations. The absence of disease in specimens of S. rotundifolia is consistent with field observations, but
its presence in natural populations of S. latifolia (especially in Virginia) suggests that the disease is recently introduced. Changes in
the host distributions were also evident. The relative abundance of S. caroliniana declined over time (especially further north), while
the relative abundance of S. virginica increased. Silene latifolia was absent or rare south of Pennsylvania before ca. 1920, indicating
that S. latifolia and its anther smut are likely to be recent introductions in Virginia. Methods are also presented that quantify the
completeness of coverage provided by herbarium specimens.
Key words: Caryophyllaceae; fungal pathogen; invasive species; Microbotryum violaceum; plant distribution; Silene antirrhina;
Silene noctiflora;Silene ovata;Silene stellata;Silene vulgaris.
Although gathering data on disease distribution is routine in
agriculture and medicine, disease distributions are far less well
understood in natural populations, largely because the insti-
tutional infrastructure for gathering such data does not exist
or is put in place only when natural populations are seen as
likely disease reservoirs (rabies: Krebs et al., 2000; West Nile
virus: Eidson et al., 2000; Hanta virus: Abbott et al., 1999).
With regard to plant populations specifically, there is almost
no source of information on disease distributions on different
plant species, other than in general catalogs (Farr et al., 1989;
Bu¨chen-Osmond et al., 2000) that are not comprehensive
enough to obtain quantitative information on the disease dis-
tribution or long-term disease trends. It has been recently ar-
gued that understanding the spatial scales of plant–pathogen
interactions is essential to explaining their dynamics (Burdon
and Thrall, 1999, 2001). For natural plant–pathogen systems
in particular, such regional studies are far fewer. Rust flax on
Linum (Jarosz and Burdon, 1991) and anther smut and other
diseases on island archipelagos (Carlsson et al., 1990; Carlsson
and Elmqvist, 1992; Ericson et al., 1999) are a few examples.
A major source of information on plant species distribution
comes from herbarium collections. While these collections
have been primarily intended to support taxonomic and floris-
tic studies, they have also been a useful source of information
on life-history traits (Primack, 1978, 1980), stomatal densities
1
Manuscript received 30 January 2003; revision accepted 1 May 2003.
The authors thank Robert Wilbur of the Duke University Herbarium for his
encouragement and help with obtaining the loans of specimens. Elizabeth
Richardson and A. M. Heuhsen assisted with data collection, entry, and anal-
ysis. Many other people helped with the fieldwork mentioned in this study,
including Deb Hlavaty, Kara O’Keefe, and Caroline Hughes. Support was
provided in part by grants from NSF (DEB 007654) and NIH (GM 6076601)
to J. A. and an NSF pre-doctoral fellowship to A. M. H.
4
E-mail: Antonovics@virginia.edu; Telephone: 434-243-5076.
(Parkhurst, 1978), phenology (Borchert, 1996), history of in-
vasive weeds (Forcella et al., 1986; Forcella and Harvey,
1988), and antibiotic effects of plant extracts (Eloff, 1999).
They are also increasingly being used to document the occur-
rence of pathogens (McCain and Hennen, 1986; Plowman et
al., 1990; Vergeer and Denhartog, 1991; Clay, 1993; Rabeler,
1993; Fraile et al., 1997; Barreto et al., 1998; Mouchacca and
Horak, 1998; Pimentel et al., 1998; Ristaino, 1998; Koponen
et al., 2000; Ristaino et al., 2001) and herbivores (Graham,
1995).
In this study, we used herbarium collections to investigate
the distribution of anther-smut disease, caused by Microbo-
tryum violaceum, on several eastern U.S. species of Silene.
Our goals were both to identify localities where the disease
was likely to be found and to obtain data on the distribution
and long-term dynamics of the disease and its hosts. Because
these herbarium specimens span time periods back into the
1800s, we were also able to investigate whether there have
been any changes in the distribution or abundance of the dis-
ease and its host species over time. We also asked if the oc-
currence of disease is related to host density or to the presence
of disease in neighboring regions.
MATERIALS AND METHODS
Study species—Microbotryum violaceum sensu lato causes anther-smut dis-
ease in over 100 species in the Caryophyllaceae and related families (Thrall
et al., 1993). The disease is characterized by the production of dark-colored
fungal spores instead of pollen in the anthers of an infected plant. Diseased
flowers also have reduced ovaries and are therefore completely sterile. The
disease is largely pollinator transmitted. Although formerly classified as Us-
tilago violacea, anther smut is now considered to be in the order Microbotry-
ales (Urediniomycetes), which is evolutionarily quite distant from the grass
smuts (Ustilago spp.) in the Ustilaginomycetes (Begerow et al., 1997). Mi-
crobotryum violaceum is itself almost certainly a species complex. Distinct
October 2003] 1523A
NTONOVICS ET AL
.—D
ISTRIBUTION OF ANTHER SMUT
host races have been established experimentally (Goldschmidt, 1928; Liro,
1938; Antonovics et al., 1996), and extensive genetic divergence has been
shown between the anther smuts on native North American species and anther
smut on the introduced S. latifolia (Perlin et al., 1996). In Europe, Microbo-
tryum isolates from different species that have morphological differences have
often been given species recognition (Vanky, 1998). Because the systematics
of the anther smut fungi is in a state of flux, we use the name M. violaceum
for all isolates regardless of the host, but with no implication that they rep-
resent a single species or host-race of the fungus.
We investigated the distribution of anther smut on four species of Silene
that have been the focus of more detailed genetic and ecological investigations
in our laboratory. Three of these species are native to North America (S.
virginica L.—fire-pink; S. caroliniana Walter—Carolina pink; and S. rotun-
difolia Nutt.—roundleaf catchfly) and one is introduced (S. latifolia Poiret—
white campion [5S. alba (Miller) Krause, 5Melandrium album (Miller)
Garcke, 5Lychnis alba Miller]).
Herbarium collections—Specimens were borrowed from the following her-
baria, chosen to cover the expected range of these species in the eastern USA
(alphabetically by herbarium code; Holmgren et al., 1990): Alabama Museum
of Natural History (ALU); Cornell University, Bailey Hortorium (BH); Ohio
University (BHO); University of Cincinnati (CINC); Clemson University
(CLEMS); Carnegie Museum of Natural History, Pittsburgh (CM); University
of Georgia (GA); Gray Herbarium, Harvard University (GH); University of
Illinois (ILL); Illinois Natural History Survey (ILLS); Northern Kentucky
University (KNK); University of Kentucky (KY); University of Michigan
(MICH); Missouri Botanical Garden (MO); North Carolina State University
(NCSC); New England Botanical Club (NEBC); New York Botanical Garden
(NY); New York State Museum, Albany (NYS); Pennsylvania State Univer-
sity (PAC); Academy of Natural Sciences, Philadelphia (PH); University of
Tennessee (TENN); University of Missouri (UMO); Smithsonian Institution,
Washington (US); United States National Herbarium (USNH); Vanderbilt Uni-
versity (VDB); Virginia Polytechnic Institute and State University (VPI); Val-
dosta State University, Georgia (VSC); West Virginia University (WVA).
Collection coverage—To assess the degree to which the collections pro-
vided an adequate description of the species distributions, we created a model
based on random resampling. Data on specimens of a species were chosen at
random without replacement, and a record kept of whether choosing an ad-
ditional specimen resulted in the recording of a new county. This random
drawing was iterated 100 times to assess the average number of new counties
found per additional specimen drawn. The following three-parameter model
was then fitted to the data: y5k1cx 1aexp(2bx), where y5the number
of new counties per additional specimen, x5the number of specimens ex-
amined, and the other parameters are estimated from the data. This model
provided the best fit of several two- and three-parameter models that were
investigated. The number of specimens that would have been required to
obtain complete coverage was obtained from the intercept of this curve on
the x-axis (i.e., when no new counties would be found). The predicted total
number of counties occupied with complete coverage was obtained by inte-
grating the function to find the area under the curve from 0 to the x-intercept.
Disease assessment—Nearly all specimens were flowering and could be
scored visually for the presence of anther-smut disease. The presence of spores
in the anthers was confirmed with a binocular microscope. Overall we examined
1022 specimens of S. virginica, 888 of S. caroliniana, 96 specimens of S.
rotundifolia, and 1104 of S. latifolia. Plant species identification was confirmed
on all specimens and the sheets annotated appropriately. Although several sub-
species have been recognized within S. caroliniana (Wilbur, 1976) we made no
attempt at subspecies identification. We noted the date of collection and the
county and state where the collection was made. We pooled duplicate sheets
(representing specimens from one collecting event at one time) that had been
distributed to several herbaria, as well as multiple collections made from the
same county on the same date (because this might have represented deliberate
intensive local search for a particular species, thus biasing the overall sampling).
We considered any collection diseased if that herbarium sheet or its duplicates
contained at least one diseased plant. Because multiple individuals were only
rarely sampled from any one locality, it was not possible to estimate disease
prevalence within individual collecting sites.
Data analysis—The data were analyzed using ArcView GIS 3.3 (Environ-
mental Systems Research Institute, 2002) and Mathematica 4.1 (Wolfram Re-
search, 2000). Because precise locality data were frequently not noted on the
labels, we assigned the coordinates of the county capital (or county seat) as
the location of a specimen collected in that county. Statistical analysis of
derived data was carried out using SAS version 8.02 (SAS Institute, 2001)
and SigmaPlot 5.00 (SPSS, 1999).
Changes in Silene abundance and geography over time—To examine
changes in the frequency of collections over time, visual maps of the geo-
graphic range were created by categorizing the data into quartiles over time,
with each quartile containing an equal number of specimens. This enabled
comparisons of distributions over time without the confounding effects of
sampling intensity.
We quantified the change in geographic range for the collections of each
Silene species by calculating the north–south and east–west distance from a
fixed point for each specimen. Then we took the average of these distances
for each decade to obtain the change in the average distribution of the col-
lections for each 10-yr period. The relative abundance of the collections of
each plant species over time was measured by counting the number of spec-
imens for each species in a decade, then expressing that number as a per-
centage of the total number of specimens for that species.
Differential changes in broad geographic patterns among the species col-
lections were examined statistically by dividing the study region into five
regions. We then used PROC FREQ in SAS to test if temporal changes in the
distribution of the collections of each species were significantly different from
each other. The following regions were defined: Northeast (Maine, New
Hampshire, Vermont, Massachusetts, Rhode Island, Connecticut), Mid-Atlan-
tic (New York, Pennsylvania, Delaware, New Jersey, Maryland, and District
of Columbia), South (Virginia, North Carolina, South Carolina, Tennessee,
Georgia, Florida, Alabama, and Mississippi), Midwest (West Virginia, Ohio,
Kentucky, Indiana, Michigan, Illinois, and Wisconsin) and West (all states
west of the Mississippi River). These regions were chosen because together
they included the distribution of all four species, and because they were, to
some extent, separable phytogeographically and historically.
Factors influencing prevalence of disease—For the two species in which
diseased specimens were present (S. caroliniana and S. virginica), we ex-
amined whether the likelihood of disease in a specimen was correlated with
either collection density (as indicated by the number of specimens of that
species collected in a county or region) or with the prevalence of disease in
the neighboring region (as indicated by the fraction of diseased specimens).
We did not correct for county area, because the variation in specimen number
was far greater than the variation in county size; there was no significant
correlation of specimen number with county size in either species.
We calculated the likelihood of disease in a specimen as the fraction of
diseased specimens in a given county. We then calculated the number of
specimens collected in that ‘‘focal’’ county, and the number collected in sur-
rounding counties within 30, 30–40, 40–50, and 50–60 miles (1 mile 5
1.6093 km). The number collected within 30 miles excludes other specimens
within the focal county. Distances from the focal county to neighboring coun-
ties were determined by calculating the distances from the seat of the focal
county to the seats of the neighboring counties. To assess if there was a
statistically significant relationship between the fraction of diseased specimens
in a focal county and the number of specimens collected in that same county
or in adjacent counties, we used logistic regression (LOGISTIC procedure in
SAS). Each specimen within a county was given a score of 0 (healthy) or 1
(diseased) to use as the dependent variable in the logistic regression. We used
the same method to establish whether the likelihood of disease in a specimen
was related to the fraction of diseased specimens in neighboring counties.
We first carried out these analyses using all counties as focal counties.
Because the vast majority of the counties had no diseased specimens, we then
1524 [Vol. 90A
MERICAN
J
OURNAL OF
B
OTANY
T
ABLE
1. Estimated coverage of the total species range for Silene caroliniana,S. latifolia,S. virginica, and S. rotundifolia, using the methods
described in the text and in Fig. 1. The estimated number of necessary sheets represents the estimated average number of herbarium sheets
that would have to be drawn in order to identify the entire range of occupied counties for a species.
Species
Occupied counties
No. observed Estimated total Percent coverage
Collections
No. sampled Estimated no.
necessary sheets Percent more
specimens needed
S. caroliniana
S. latifolia
S. virginica
S. rotundifolia
230
421
402
45
246.0
508.9
480.3
91.9
93.50
82.70
83.70
49.00
888
1104
1022
96
1306.1
2036.9
1842.9
531.0
47.10
84.40
80.20
453.10
Fig. 1. The likelihood of discovering a specimen in a new county for each
additional herbarium sheet chosen, for the four Silene species studied, and for
diseased S. virginica. Each data point reflects the average of 100 random
selections, given the actual recorded number of specimens and occupied coun-
ties for each species. The fitted curves are of the form y5k1cx 1
aexp(2bx).
restricted the analyses to include as focal counties only those having disease
specimens at some time during the study. To eliminate the ‘‘propositus ef-
fect,’’ we estimated the frequency of disease in a focal county as (d21)/(n
21), where d5the number of diseased plants and n5the total number of
plants. This frequency takes into account the fact that there has to be at least
one diseased individual in a population for it to be recognized as diseased;
failing to make this correction would result in an automatic expectation that
counties with a few samples would have a higher disease frequency.
RESULTS
Collection intensity—Although extrapolating curves be-
yond the actual data is always to be interpreted cautiously, our
analysis (Table 1; Fig. 1) suggested that we had identified over
80% of the counties that were occupied by the three most
abundant species (S. caroliniana,S. virginica, and S. latifolia),
and ca. 50% of the counties occupied by S. rotundifolia. The
model also showed that considerably more sampling effort
would be needed to obtain complete coverage for these spe-
cies. Thus, in S. caroliniana we would have to examine half
as many herbarium specimens again to find all the counties,
while in S. virginica and S. latifolia about 80% more samples
would be needed. In S. rotundifolia almost five times as many
specimens would be needed to identify all the counties.
Disease distribution—Silene virginica—In S. virginica, an-
ther-smut disease was found in 85 of 1022 collections. Dis-
eased collections were present throughout the species range
(Fig. 2a). Analysis of disease frequency in 20-yr intervals in-
dicated that the frequency of herbarium specimens that are
diseased has increased from ca. 3.2% before 1900 to ca. 14.9%
for specimens collected after 1980 (Fig. 3).
The fraction of diseased specimens in a county was not
significantly related to the density of specimens in that county,
either when all counties were included in the analysis (regres-
sion coefficient 520.0008; P50.40) or when the analysis
included only those counties with disease (regression coeffi-
cient 520.0009; P50.21). In all cases, year was used as a
covariate and had a highly significant effect.
The fraction of diseased specimens in a county was also not
significantly related to the number of specimens within 0–30,
30–40, 40–50, and 50–60 miles when all counties were in-
cluded as focal counties in the analyses (Pvalues for effect of
neighbor distance classes 30–60 miles ranged from 0.32 to
0.86; 1 mile 51.6093 km). However, when only those focal
counties with diseased specimens were used in the analysis,
the logistic regression of disease vs. healthy on year and num-
ber of neighbors in the different distance classes was highly
significant (P,0.0001); both year and distance class contrib-
uted significantly to the model. Selective elimination of the
furthest distance classes showed no significant contributions
of the numbers at 50–60 miles, while the effects of the other
three distances were significantly negative. Their relative ef-
fects depended on how many variables were included in the
model (see Table 2). All year by distance interactions were not
significant, regardless of the number of distance classes en-
tered into the model. Therefore, there is evidence that counties
that are more isolated (i.e., have fewer close neighbors) have
a higher prevalence of disease.
There was no significant relationship between the fraction
of diseased specimens in a given county and the fraction of
diseased specimens in neighboring counties, although the trend
was for more disease within the immediate neighboring area
(i.e., 0–30 miles) (logistic regression for all counties 50.84,
P50.08; logistic regression when only diseased counties are
included 50.52, P50.37; year included in the model).
Because we had a relatively large number of diseased spec-
imens of S. virginica, we also examined whether the collec-
tions provided an adequate description of the disease distri-
butions in a manner analogous to our analysis for the host
distributions (see Materials and Methods: Collection cover-
age). The analysis showed that we were far from attaining
complete coverage for the disease (Fig. 1). Almost every new
October 2003] 1525A
NTONOVICS ET AL
.—D
ISTRIBUTION OF ANTHER SMUT
Fig. 2. The distribution of collections of four Silene species in the eastern United States. Diseased specimens are shown as black dots, and healthy specimens
are shown as gray dots. Each dot is placed in a randomly generated location within the county of the collection’s origin in order to minimize overlapping points.
(a) Silene virginica. (b) S. caroliniana. (c) S. latifolia. (d) S. rotundifolia.
Fig. 3. The prevalence of disease in Silene virginica and S. caroliniana
over 20-yr periods. Straight lines are regressions of prevalence vs. time
weighted by number of specimens at each time (as indicated).
T
ABLE
2. Effect of year and number of specimens of Silene virginica
sampled in adjacent counties on the likelihood of disease on a
specimen in a focal county (including only those counties where
diseased specimens had been found). The model estimates are
based on multivariate logistic regression. One mile
5
1.6093 km.
Effect Slope
estimate Chi
square P
Four-variable model
Year
0–30 miles
30–40 miles
40–50 miles
0.018
2
0.01
2
0.022
2
0.038
9.01
0.77
3.68
6.69
0.0027
0.38
0.055
0.0097
Three-variable model
Year
0–30 miles
30–40 miles
0.016
2
0.026
2
0.031
8.06
6.42
9.08
0.0045
0.011
0.0026
Two-variable model
Year
0–30 miles 0.019
2
0.031 13.36
8.93 0.0003
0.0028
1526 [Vol. 90A
MERICAN
J
OURNAL OF
B
OTANY
Fig. 4. The relative abundance of collections of Silene caroliniana,S.
latifolia, and S. virginica over 10-yr intervals. Each data point represents the
percentage of all the specimens for the particular species that was collected
within the decade. See Fig. 8 for key to symbols.
T
ABLE
3. Comparison of geographical changes according to species
of Silene. The data represent the number of collections for each
species in each region for four equally divided temporal quartiles
(see Fig. 5 for discussion of temporal quartiles).
Species
Quartiles
1234
Total of
region Percent of
total
New England, All
S. latifolia
S. caroliniana
S. virginica
161
103
58
0
29
26
3
0
13
12
1
0
19
18
1
0
222 7.37
Mid-Atlantic, All
S. latifolia
S. caroliniana
S. virginica
386
100
237
49
260
147
85
28
322
203
84
35
122
72
25
25
1090 36.16
Midwest, All
S. latifolia
S. caroliniana
S. virginica
71
24
9
38
149
49
40
60
157
79
30
48
208
81
30
97
585 19.41
South, All
S. latifolia
S. caroliniana
S. virginica
103
3
33
67
275
37
85
153
224
34
71
119
354
50
79
225
956 31.72
West, All
S. latifolia
S. caroliniana
S. virginica
32
21
3
8
47
22
8
17
33
10
1
22
49
13
5
31
161 5.34
Total of quartiles
Percent of total 753
24.98 760
25.22 749
24.85 752
24.95 3014 100
collection was from a new county (there were a total of 85
collections from 73 counties). Because of the need to extrap-
olate well beyond the data, estimates of the percentage of
counties in which the disease might be expected to occur dif-
fered greatly between replicate runs (each with 100 randomi-
zations) but were generally between 40% and 100% of the
total counties currently occupied by the host.
Silene caroliniana—In S. caroliniana, anther-smut disease
was found in 25 of 888 collections (Fig. 2b). Analysis of dis-
ease frequency in 20-yr intervals indicated that the frequency
of herbarium specimens that are diseased has increased sig-
nificantly from ca. 0.6% before 1900 to ca. 6.7% for speci-
mens collected after 1980 (Fig. 3). Silene caroliniana has a
highly disjunct distribution, but the disease was found in spec-
imens from all these disjunct regions. Three subspecies of S.
caroliniana have been recognized (Wilbur, 1976). Silene car-
oliniana subsp. pennsylvanica occurs in all the eastern states
from North Carolina to New England. Silene caroliniana
subsp. caroliniana includes the plants in South Carolina, Geor-
gia, and southern North Carolina. Silene caroliniana subsp.
wherryi includes the populations located in Kentucky/Ohio,
Tennessee, and Alabama/Georgia. These subspecies have been
distinguished on the basis of detailed morphological characters
(Wilbur, 1976). Using geographic location to define the sub-
species, we found that anther-smut disease was present in all
of the subspecies. However, disease frequency among the three
subspecies was significantly different (disease frequency for S.
c. pennsylvanica 51.44%, n5693; S. c. caroliniana 5
4.49%, n589; S. c. wherryi 510.58%, n5104; P,
0.0001). Pairwise tests showed the overall significance was
largely due to S. c. wherryi and S. c. caroliniana specimens
having higher disease levels than S. c. pennsylvanica. The dif-
ference between S. c. caroliniana and S. c. wherryi was not
significant, P50.17.
We also investigated disease relationships in the same way
as in S. virginica, but almost none of the distance-related ef-
fects were significant, perhaps because the sample size of dis-
eased specimens was much smaller than in S. virginica. With
all counties included, the fraction of diseased specimens in a
focal county correlated positively with the percentage fraction
of disease in neighboring counties, although the only signifi-
cant values were for neighbors within 30–40 miles (logistic
regression 50.089, P50.074) and 40–50 miles (logistic
regression 50.042, P50.028). When the analysis included
only those focal counties with disease, the trend was similar
but no values were significant (Pvalues for effect of neighbor
distance classes 50.36–0.80).
S. latifolia and S. rotundifolia—No disease was found in the
collections of these species (Fig. 2c, d).
Host distribution—Collections of the three most abundant
species were most frequent between 1930 and 1960 (Fig. 4).
Silene caroliniana was the most commonly collected species
before 1910, while S. virginica was most commonly collected
after 1960. The relative changes in collection frequencies were
highly significant (P,0.0001 for year 3species interaction,
logistic regression). To examine whether these changes were
different in different regions, we divided the overall number
of collections into four time periods, each with an approxi-
mately equal number of collections (i.e., quartiles), and ex-
amined whether the relative abundance of the species differed
among regions over the time quartiles (Table 3). The results
had a highly significant interaction for region 3quartile 3
species (P,0.0001, Proc Catmod in SAS), showing that the
relative abundance of the species also changed differently in
the different regions.
Visual inspection suggested that there had been substantial
change in the geographical range of the collections of the three
more abundant species. In S. latifolia, collections were largely
restricted to the Northeast and Mid-Atlantic regions before
1930 (Fig. 5a, b). After that, the collections become more
abundant both in the South and Midwest regions.
In S. virginica, the overall collection distribution remained
unchanged, but there was an increase in the frequency of col-
lections in the South and West regions (Fig. 6a, b).
In S. caroliniana, the early collections were mainly from
October 2003] 1527A
NTONOVICS ET AL
.—D
ISTRIBUTION OF ANTHER SMUT
the Mid-Atlantic region, while later collections extended its
distribution to Ohio, Kentucky, and Missouri (Fig. 7a, b).
There was a substantial increase in the number of collections
in the South and a decrease in the number of collections in
the Mid-Atlantic region. Statistical analysis of average dis-
tance of the collections from a set point confirmed the trend
of increasing numbers of collections towards the south and
west for all three species (Fig. 8).
For S. rotundifolia, there was no evidence of change in
abundance or distribution with time (Fig. 2d).
DISCUSSION
Disease distribution—Our study shows that it is possible to
identify herbarium specimens of Silene that are diseased with
anther smut and that such specimens provide information on
temporal and spatial patterns of both host and pathogen dis-
tribution. Thus, our data show that the incidence of anther-
smut disease in both S. virginica and S. caroliniana has in-
creased over the previous century. It is very unlikely that these
results are caused by any increase in tendency to collect dis-
eased specimens. Indeed, we could find no evidence that dis-
eased specimens were recognized by collectors, let alone that
the propensity to collect such specimens had increased with
time. We only found one herbarium sheet (of S. virginica) that
was annotated to indicate the diseased specimens were in any
way unusual, and then it was not recognized as diseased—the
label simply noted that the plants had ‘‘purple stamens.’’ In S.
virginica, the disease is relatively inconspicuous because the
flowers have deep red petals and dark-colored anther sacs. In-
deed, the disease was first noted on S. virginica in 1988 in a
natural roadside population (Antonovics et al., 1996), even
though it was collected more than 100 years ago! In S. caro-
liniana, diseased plants are much more easily recognizable by
their dark anthers, and this may have introduced some bias.
However, a systematist who had extensively examined both
field and herbarium material of this species (Wilbur, 1976) had
never noticed the disease (R. Wilbur, Duke University, per-
sonal communication), suggesting that collection bias by other
investigators might have been minimal.
Explaining the reason for the increase in anther smut on
these species is more difficult. In both S. virginica and S. car-
oliniana, diseased specimens were more likely to have other
diseased specimens in neighboring regions, suggesting an ‘‘ep-
idemiological signal’’ in the data; however, this relationship
was only significant in S. caroliniana. There was no evidence
that counties where more plant specimens had been collected
had more disease. This suggests that counties with more spec-
imens (and therefore possibly more populations) do not result
in greater disease incidence. Rather the trend was in the op-
posite direction in S. virginica in that there was a highly sig-
nificant negative relationship between disease incidence and
the overall number of specimens collected in the immediately
surrounding counties. In S. caroliniana, the trend was in the
same direction, but it did not approach significance. One ex-
planation for this pattern is that disease might be more com-
mon in recently established and/or smaller populations.
Our results show that all three subspecies of S. caroliniana
are susceptible to the disease, and the pathogen occurs in high-
ly disjunct regions of the host. This supports the idea that this
host–pathogen association is a very old one. It is interesting
that S. c. subsp. pennsylvanica has a lower disease incidence
than the other two subspecies, but the reason for thisis unclear.
Transfer of the disease between S. caroliniana and S. virginica
cannot be excluded. We have found diseased populations of
each species within several hundred meters of each other, and
although S. caroliniana generally flowers earlier in the season,
there is substantial overlap in their flowering times. The two
species are also known to hybridize (Mitchell and Uttal,1969).
We have found two obviously hybrid populations in the field,
but neither was diseased with anther smut. We therefore do
not know if such hybrids might form a bridge through which
host shifts might occur. The issue of whether Microbotryum
represents one or two host races on these two native North
American species needs to be resolved experimentally and
phylogenetically.
No diseased plants were seen in the collections of S. rotun-
difolia. Far fewer collections of this species were available,
but no disease has been seen in numerous populations that
were visited in the field as part of a study investigating genetic
differentiation in this species (Leonie Moyle, University of
California, Davis, personal communication). Silene rotundi-
folia is a habitat specialist found almost exclusively on sand-
stone bluffs in West Virginia and Tennessee. Although it is
often close to populations of S. virginica and also has bright
red flowers that are hummingbird pollinated, it flowers in Au-
gust, much later than S. virginica. There have been no studies
of the susceptibility of S. rotundifolia to anther-smut disease.
Because our own research has been focused largely on an-
ther smut on S. latifolia, we examined numerous herbarium
specimens of this introduced species in an attempt to under-
stand its disease history. We were therefore surprised that we
found no diseased specimens of S. latifolia in over 1000 her-
barium sheets that we examined. This may be due to collection
bias (the disease is particularly conspicuous on S. latifolia be-
cause of the white flowers and because female plants produce
diseased anthers and have aborted ovaries). Alternatively, the
disease on this species may be a relatively recent introduction
from Europe. We know from field studies that diseased S. la-
tifolia are commonly found in the mountains and western re-
gions of Virginia (Antonovics et al., 1996), and recent studies
have recorded it occasionally in Pennsylvania (A. M. Jarosz
and E. Lyons, Michigan State University, personal communi-
cation) and Michigan (J. Antonovics, University of Virginia,
personal observation). The disease has not been found in New
England, but does occur on Nantucket Island (T. Meagher,
University of St. Andrews, UK, personal communication). One
possibility that might explain the present-day distribution of
disease is that M. violaceum has ‘‘followed’’ the relatively re-
cent spread of S. latifolia into the southeastern USA and that
it has been largely lost in areas where S. latifolia has been
established for a long time. Therefore, in Virginia, where most
of our studies have been carried out, both the host and the
disease are probably recent introductions. Previous studies
have shown that the anther-smut fungus on S. latifolia is a
host-race distinct from that found on S. virginica (Antonovics
et al., 1996) and S. caroliniana (Perlin et al., 1996). This study
confirms that the disease on S. caroliniana and S. virginica
could not have come from the introduced S. latifolia, because
disease was present on the native species well before the
spread of S. latifolia into regions of sympatry.
Several other introduced and native species of Silene occur
in the eastern United States. With the exception of S. vulgaris,
we have not seen anther smut on any of the introduced species.
In Virginia, we have frequently observed S. noctiflora L. and
S. antirrhina L. sympatric with diseased S. latifolia. However,
1528 [Vol. 90A
MERICAN
J
OURNAL OF
B
OTANY
Figs. 5–7. The difference in geographic range in the first temporal quartile
of collections, top, and in the fourth quartile, bottom. The quartiles consist of
an equal number of collections, rounded to the year. The dots represent oc-
cupied counties, and the size of the dots reflects the number of collections
made in that county. The ranges were created in ArcView by making contour
lines using inverse distance weighting (12 nearest neighbors, power of two).
The shaded areas are bounded by the halfway distance between the contours
representing zero and one plant per county, but with a regional weighting
such that most isolated points are excluded. This method has some bias be-
cause it overestimates the range near borders and near large counties. In all
maps due north is toward the top of the figure. Fig. 5 (above). Silene latifolia.
The first quartile (a) contains collections made before 1930, and the fourth
quartile (b) contains collections made from 1958 to the present.
Fig. 6. Silene virginica. The first quartile (a) contains collections made
before 1934, and the fourth quartile (b) contains collections made from 1970
to the present.
these species are annuals and disease persistence on them is
unlikely (Thrall et al., 1993). Moreover, the flowers of S. an-
tirrhina are small and inconspicuous, and this species is almost
certainly highly self-pollinated and/or partially cleistogamous.
Silene vulgaris (Moench) Garcke is perennial, and while pop-
ulations are generally free of anther smut, we have found a
population with diseased plants of S. vulgaris growing with
heavily diseased S. latifolia. This represents a recent and local
host-shift (Antonovics et al., 2002) and a similar host-shift has
been observed in Europe (Hood et al., 2003).
Two other native species of Silene occur in the easternUnit-
ed States. Silene stellata (L.) Ait. is quite common, but we
have never found anther smut on this species. Silene ovata
Pursh. is a much rarer species restricted to relatively fewforest
glade sites in the southern Appalachians. Although it has not
been observed diseased in nature, a few plants did become
accidentally diseased with anther smut when growing in pots
near diseased S. latifolia in a greenhouse (L. Moyle, Univer-
sity of California, Davis, personal communication). This sug-
gests that S. ovata populations may be very susceptible to
anther smut but remain disease free because they are spatially
and temporally isolated.
Host distribution—Although we gathered the herbarium
data primarily to assess disease incidence, this data also pro-
vides interesting information about host distribution. Above
all, it points to the potential usefulness of herbarium data in
establishing the current and historical ranges of plant species
and in indicating long-term changes in distributions that are
useful indicators of habitat change, species decline (Meagher
October 2003] 1529A
NTONOVICS ET AL
.—D
ISTRIBUTION OF ANTHER SMUT
Fig. 7. Silene caroliniana. The first quartile (a) contains collections made
before 1907, and the fourth quartile (b) contains collections made from 1955
to the present.
Fig. 8. The average latitude and longitude of Silene latifolia,S. caroli-
niana, and S. virginica over 10-yr intervals, expressed as miles north and east
of a fixed point (specifically Charlottesville, Virginia; 1 mile 51.6093 km).
All trendlines are linear and weighted according to the number of specimens
per decade.
et al., 1978; Kiang et al., 1979), or invasions (Forcella et al.,
1986; Forcella and Harvey, 1988). Such long-term studies
have been routine in many European countries with a strong
tradition of recording and collection (Preston et al., 2002), but
they are much rarer in the United States.
Our studies show the enormous potential of herbarium data
in delimiting the distribution of species in the United States.
Even though we obtained specimens from only 28 herbaria,
the effectiveness of the coverage was very high. Thus, in the
three most abundant species, we estimated that we had iden-
tified well over 80% of the counties in which the species were
likely to occur. Adding between 500 and 1000 samples (if the
specimens actually exist) would result in close to 100% cov-
erage. Clearly these figures are very approximate, as they do
not take into account ongoing changes in plant distribution or
regional biases in herbarium collections. To examine this we
compared our coverage with the coverage as indicated in pub-
lished county level distribution maps for Pennsylvania, Vir-
ginia, North Carolina, and South Carolina. There was clear
evidence of ‘‘herbarium bias.’’ For example, in Pennsylvania
and South Carolina, our county coverage corresponded well
(between 82.0 and 91.94%, averaged over all species except
S. rotundifolia) with the published distribution because we
sampled the major herbaria in these states. However, in Vir-
ginia and North Carolina our coverage was poor (between 51.0
and 54.7%). We were unaware at the time of the substantial
collections at Longwood College in Virginia, and we inadver-
tently did not sample the University of North Carolina her-
barium because we moved institutions halfway through the
study.
Indeed, it would be interesting to quantify the incremental
contribution of added samples within herbaria vs. added sam-
ples from different herbaria. Like species diversity in com-
munities, we can envisage that the diversity of herbarium col-
lections could be partitioned into categories analogous to those
used by community ecologists, namely alpha (or within-her-
barium) and beta (between herbarium) diversity. Although we
have not done this, our overall analyses point to the fact that
it is possible to quantify to a remarkable degree the adequacy
of herbarium sampling in describing the distribution of any
particular species. In contrast to the host species, our results
for anther smut show that the effective sampling of the disease
has been very inadequate because adding more specimens of
the host still leads to a high probability of detecting counties
where the disease has not yet been discovered (see Fig. 1, M.
violaceum on S. virginica).
This study shows large changes in the distribution of the
collections of three of the four species that we studied (in S.
rotundifolia the number of collections was too few to discern
any clear trends). The main question that arises is whether
these changes represent collection bias or whether they truly
reflect changes in plant distribution. Thus, the increase over
time in the number of collections made toward the south and
west is consistent not only with the idea that human distur-
bance and movement in these areas occurred later than in New
England and the Mid-Atlantic states, but also with the fact that
there has been a greater increase in academic institutions in
these areas over the past century. Prather et al. (2003) also
found that herbaria in the southeastern United States started
collections later than did those in the northeast and that their
period of peak collection was also much later. Their results
were generally consistent with ours with regard to overall col-
1530 [Vol. 90A
MERICAN
J
OURNAL OF
B
OTANY
lecting frequency, which peaked between 1930 and 1950 (even
though we used different focal species and fewer herbaria in
our study).
However, real changes in species abundance are likely to
have taken place, as judged from the relative frequency of the
collections of the species that we have studied. The decrease
in the abundance of S. caroliniana collections relative to the
number of collections of the other common species strongly
suggests that the abundance of this species may be declining,
especially in New England and the Mid-Atlantic. This is con-
sistent with our own natural history observations that the pop-
ulations are often highly fragmented, very small, and absent
in spite of good locality data from herbarium records. Our
studies also show clearly that S. latifolia was largely confined
to New England in the earlier part of this century but then
spread southwards. The first collection of S. latifolia further
south than Pennsylvania or New Jersey was in 1924, and there
were no collections from Kentucky until 1950. Given the large
number of collections and the fact that flowering plants are
very conspicuous, the earlier absence of S. latifolia in the
southern and western United States is unlikely to be due to
collecting bias. Silene latifolia is itself introduced into the
United States from Europe (McNeill, 1977), and our studies
indicate that this introduction was initially in the New England
region (or perhaps Canada).
A number of previous studies have emphasized the impor-
tance of herbarium specimens in understanding not just species
distributions but also other important aspects of plant biology.
In this study, we have shown that herbarium specimens can
be valuable in showing changes in both disease abundance and
host distribution. We have also shown that it is possible to
assess the completeness (or precision) of the coverage provid-
ed by such data and to approach the problem of collection bias
(or accuracy) by focusing on relative rather than absolute
abundances of collections. The advent of computer-based im-
aging and record keeping is likely to greatly increase acces-
sibility and hence the usefulness of herbarium data for these
types of study in the future.
LITERATURE CITED
A
BBOTT
, K. D., T. G. K
SIAZEK
,
AND
J. N. M
ILLS
. 1999. Long-term hantavirus
persistence in rodent populations in central Arizona. Emerging Infectious
Diseases 5: 102–112.
A
NTONOVICS
, J., M. E. H
OOD
,
AND
J. P
ARTAIN
, 2002. The ecology and ge-
netics of a host-shift: Microboryum as a model system. American Nat-
uralist 160: S40–S53.
A
NTONOVICS
, J., D. S
TRATTON
,P.H.T
HRALL
,
AND
A. M. J
AROSZ
. 1996. An
anther-smut disease (Ustilago violacea) of fire-pink (Silene virginica): its
biology and relationship to the anther-smut disease of white campion
(Silene latifolia). American Midland Naturalist 135: 130–143.
B
ARRETO
, R. W., H. C. E
VANS
,
AND
R. E. H
ANANDA
. 1998. First record of
Cercospora pistiae causing leaf spot of water lettuce (Pistia stratioites)
in Brazil, with special reference to weed biocontrol. Mycopathologia
144: 81–85.
B
EGEROW
, D., R. B
AUER
,
AND
F. O
BERWINKLER
. 1997. Phylogenetic studies
on nuclear large subunit ribosomal DNA sequences in the smut fungi
and related taxa. Canadian Journal of Botany 75: 2045–2056.
B
ORCHERT
, R. 1996. Phenology and flowering periodicity of Neotropical dry
forest species: evidence from herbarium collections. Journal of Tropical
Ecology 12: 65–80.
B
U
¨CHEN
-O
SMOND
, C., L. B
LAINE
,
AND
M. C. H
ORZINEK
. 2000. The universal
virus database of ICTV (ICTVdB). In M. H. V. van Regenmortel et al.
[eds.], Virus taxonomy. Classification and nomenclature of viruses, sev-
enth report of the International Committee on Taxonomy, 19–24. Aca-
demic Press, New York, New York, USA.
B
URDON
, J. J.,
AND
P. H. T
HRALL
. 1999. Spatial and temporal patterns in
coevolving plant and pathogen associations. American Naturalist 153:
S15–S33.
B
URDON
,J.J.,
AND
P. H. T
HRALL
. 2001. The demography and genetics of
host–pathogen interactions. In J. Silvertown and J. Antonovics [eds.],
Integrating ecology and evolution in a spatial context. Fourteenth Special
Symposium of the British Ecological Society, 197–217. Blackwell Sci-
ence, Oxford, UK.
C
ARLSSON
, U.,
AND
T. E
LMQVIST
. 1992. Epidemiology of anther-smut disease
(Microbotryum violaceum) and numeric regulation of populations of Sil-
ene dioica.Oecologia 90: 509–517.
C
ARLSSON
, U., T. E
LMQVIST
,A.W
ENNSTROM
,
AND
L. E
RICSON
. 1990. In-
fection by pathogens and population age of host plants. Journal of Ecol-
ogy 78: 1094–1105.
C
LAY
, K. 1993. The ecology and evolution of endophytes. Agriculture Eco-
systems and Environment 44: 39–64.
E
IDSON
, M., J. M
ILLER
,L.K
RAMER
,B.C
HERRY
,
AND
Y. H
AGIWARA
. 2000.
Dead crow densities and human cases of West Nile Virus, New York
State, 2000. Emerging Infectious Diseases 7: 662–664.
E
LOFF
, J. N. 1999. It is possible to use herbarium specimens to screen for
antibacterial components in some plants. Journal of Ethnopharmacology
67: 355–360.
E
NVIRONMENTAL
S
YSTEMS
R
ESEARCH
I
NSTITUTE
. 2002. ArcView GIS. Ver-
sion 3.3. Redlands, California, USA.
E
RICSON
, L., J. J. B
URDON
,
AND
W. J. M
ULLER
. 1999. Spatial and temporal
dynamics of epidemics of the rust fungus Uromyces valerianae on pop-
ulations of its host, Valeriana salina.Journal of Ecology 87: 649–658.
F
ARR
, D. F., G. F. B
ILLS
,G.P.C
HAMURIS
,
AND
A. Y. R
OSSMAN
. 1989. Fungi
on plants and plant products in the United States. American Phytopath-
ological Society, St. Paul, Minnesota, USA.
F
ORCELLA
,F.,
AND
S. J. H
ARVEY
. 1988. Patterns of weed migration in north-
western U.S.A. Weed Science 36: 194–201.
F
ORCELLA
, F., J. T. W
OOD
,
AND
S. P. D
ILLON
. 1986. Characteristics distin-
guishing invasive weeds within Echium (Bugloss). Weed Research 26:
351–364.
F
RAILE
, A., F. E
SCRIU
,M.A.A
RANDA
,J.M.M
ALPICA
,A.J.G
IBBS
,
AND
F.
G
ARCIA
-A
RENAL
. 1997. A century of tobamovirus evolution in an Aus-
tralian population of Nicotiana glauca.Journal of Virology 71: 8316–
8320.
G
OLDSCHMIDT
, V. 1928. Vererbungsversuche mit den biologischen Arten des
Antherenbrandes (Ustilago violacea Pers.). Zeitschrift fur Botanik 21: 1–
90.
G
RAHAM
, S. A. 1995. Gall makers on flowers of Cuphea (Lythraceae). Bio-
tropica 27: 461–467.
H
OLMGREN
, P. K., N. H. H
OLMGREN
,
AND
L. C. B
ARNETT
. 1990. Index Her-
bariorum, part I, 8th ed. Regnum Vegetabile 120: 1–693.
H
OOD
, M. E., J. A
NTONOVICS
,
AND
H. H
EISHMAN
. 2003. Karyotypic simi-
larity identifies multiple host-shifts of a pathogenic fungus in natural
populations. Infection, Genetics, and Evolution 2: 167–172.
J
AROSZ
,A.M.,
AND
J. J. B
URDON
. 1991. Host-pathogen interactions in nat-
ural populations of Linum marginale and Melampsora lini: II. Local and
regional variation in patterns of resistance and racial structure. Evolution
45: 1618–1627.
K
IANG
, Y. T., J. A
NTONOVICS
,
AND
L. W
U
. 1979. The extinction of wild rice
(Oryza perennis formosana) in Taiwan. Journal of Asian Ecology 1: 1–9.
K
OPONEN
, H., S. H
ELLQVIST
,H.L
INDQVIST
-K
REUZE
,U.B
ANG
,
AND
J. P. T.
V
ALKONEN
. 2000. Occurrence of Peronospora sparsa (P. rubi) on cul-
tivated and wild Rubus species in Finland and Sweden. Annals of Applied
Biology 137: 107–112.
K
REBS
, J. W., C. E. R
UPPRECHT
,
AND
J. E. C
HILDS
. 2000. Rabies surveillance
in the United States during 1999. Journal of the American Veterinary
Medical Association 217: 1799–811.
L
IRO
, J. I. 1938. Die Ustilagineen Finlands II. Annales Academiae Scientarum
Fennicae Series A 42: 1–720.
M
C
C
AIN
,J.W.,
AND
J. F. H
ENNEN
. 1986. Collection of plant material dam-
aged by pathogens: an expression of support. Taxon 35: 119–121.
M
C
N
EILL
, J. 1977. The biology of Canadian weeds. 25. Silene latifolia (Mill-
er) E. H. L. Krause. Canadian Journal of Plant Science 57: 1103–1114.
M
EAGHER
, T. R., J. A
NTONOVICS
,
AND
R. P
RIMACK
. 1978. Experimental eco-
logical genetics in Plantago. III. Genetic variation and demography in
relation to survival of Plantago cordata, a rare species. Biological Con-
servation 14: 243–257.
M
ITCHELL
,R.S.,
AND
L. J. U
TTAL
. 1969. Natural hybridization in Virginia
October 2003] 1531A
NTONOVICS ET AL
.—D
ISTRIBUTION OF ANTHER SMUT
Silene (Caryophyllaceae). Bulletin of the Torrey Botanical Club 96: 544–
549.
M
OUCHACCA
, J.,
AND
E. H
ORAK
. 1998. Annotated checklist of New Cale-
donian Basidiomycota. II. Rusts and smuts. Mycotaxon 69: 13–30.
P
ARKHURST
, D. F. 1978. The adaptive significance of stomatal occurrence on
one or both surfaces of leaves. Journal of Ecology 66: 367–383.
P
ERLIN
, M. H., C. H
UGHES
,J.W
ELCH
,S.A
KKARAJU
,D.S
TEINECKER
,A.
K
UMAR
,B.S
MITH
,S.S.G
ARR
,S.A.B
ROWN
,
AND
T. A
NDOM
. 1996.
Molecular approaches to differentiate subpopulations or formae speciales
of the fungal phytopathogen Microbotryum violaceum.International
Journal of Plant Sciences 158: 568–574.
P
IMENTEL
, G., L. M. C
ARRIS
,L.L
EVY
,
AND
R. J. M
EYER
. 1998. Genetic
variability among isolates of Tilletia barclayana,T. indica and allied
species. Mycologia 90: 1017–1027.
P
LOWMAN
, T. C., A. L
EUCHTMANN
,C.B
LANEY
,
AND
K. C
LAY
. 1990. Sig-
nificance of the fungus Balansia cyperi infecting medicinal species of
Cyperus (Cyperaceae) from Amazonia. Economic Botany 44: 452–462.
P
RATHER
, L. A., O. A
LVAREZ
-F
UENTES
,M.H.M
AYFIELD
,
AND
C. J. F
ER
-
GUSON
. 2003. The decline of plant collecting in the United States: a
threat to the infrastructure of biodiversity studies. Systematic Botany (in
press).
P
RESTON
, C. D., D. A. P
EARMAN
,
AND
T. D. D
INES
. 2002. New Atlas of the
British and Irish Flora. Oxford University Press, Oxford, UK.
P
RIMACK
, R. B. 1978. Regulation of seed yield in Plantago.Journal of Ecol-
ogy 66: 835–847.
P
RIMACK
, R. B. 1980. Phenotypic variation of rare and widespread species
of Plantago.Rhodora 82: 87–95.
R
ABELER
, R. K. 1993. The occurrence of anther smut, Ustilago violacea S.
L., on Stellaria borealis (Caryophyllaceae) in North America. Contri-
butions of the University of Michigan Herbarium 19: 165–169.
R
ISTAINO
, J. B. 1998. The importance of archival and herbarium materials in
understanding the role of oospores in late blight epidemics of the past.
Phytopathology 88: 1120–1130.
R
ISTAINO
, J. B., C. T. G
ROVES
,
AND
G. R. P
ARRA
. 2001. PCR amplification
of the Irish potato famine pathogen from historic specimens. Nature 411:
695–697.
SAS I
NSTITUTE
. 2001. The SAS system for Windows, release 8.02. SAS
Institute, Cary, North Carolina, USA.
SPSS 1999. SigmaPlot. Version 5.00. Chicago, Illinois.
T
HRALL
, P. H., A. B
IERE
,
AND
J. A
NTONOVICS
. 1993. Plant life history and
disease susceptibility: the occurrence of Ustilago violacea on different
species within the Caryophyllaceae. Journal of Ecology 81: 489–498.
V
ANKY
, K. 1998. The genus Microbotryum (smut fungi). Mycotaxon 67: 33–
60.
V
ERGEER
,L.H.T.,
AND
C. D
ENHARTOG
. 1991. Occurrence of wasting disease
in Zostera noltii.Aquatic Botany 40: 155–163.
W
ILBUR
, R. L. 1976. Infraspecific classification in the Carolina flora. Rhodora
72: 51–65.
W
OLFRAM
R
ESEARCH
. 2000. Mathematica. Version 4.1. Champaign, Illinois,
USA.