INTRODUCED LESSER CELANDINE (RANUNCULUS FICARIA, RANUNCULACEAE) AND ITS PUTATIVE SUBSPECIES IN THE UNITED STATES: A MORPHOMETRIC ANALYSIS

2009·13.5 MB·English

by  Angela R Post

INTRODUCED LESSER CELANDINE (RANUNCULUS
FICAR1A,
AND
RANUNCULACEAE)
PUTATIVE
SUBSPECIES
ITS
THE UNITED A MORPHOMETRIC
IN STATES: ANALYSIS
Angela
Post Alexander
1 Krings
R.
Department ofHorticultural Science Herbarium, Department ofPlant Biology
North Carolina State University North Carolina State University
Raleigh, North Carolina 27695-7609, USA. Raleigh, North Carolina 27695-76 U.SA.
12,
[email protected]
C
Wade
Joseph
A. Wall Neal
Department ofPlant Biology Department of
Horticultural Science
North
Carolina State University North Carolina State University
Raleigh, North Carolina 27695-76 USA. Raleigh, North Carolina 27695-7609, U.SA.
12,
ABSTRACT
iero fue introducida en los Estados Unidos como ornamental en el siglo XIX. Despues de
1;
subespecifico, nuestros objetivos en este estudio fueron determinar si existen y cuantas enti
&
Markham
Ranunculus (Ranunculaceae) Europe
ficaria L. is native to (Tutin 1964; Taylor 1978; Sell 1994;
Whittemore was
1997), but introduced to the United States (U.S.) through the garden ornamental trade
for
showy
its flowers (Bailey 1935). was collected with certainty in the U.S. in 1867 (Philadelphia County
It
Pennsylvania, Burke PH) and required 141 years to "spread" from Philadelphia, Pennsylvania Fort
s.n., to
TEX)—
Worth, Texas (Nesom FW08-1, BRIT, MO, NCSC, NCU, southernmost
the extent of the current
known
(Nesom
distribution 2008).
194
whose
In Europe, five subspecies of Ranunculus ficaria art recognized (Tutin 1964; Sell 1994), "ecol-
ogy and distribution appear to overlap, but tend to be different" (Sell 1994). Flora Europaea currently
[. .]
.
from western Europe, eastward southern subsp.
Ranunculus subsp. to Italy, (2)
recognizes: ficaria ficaria
(1)
Lambinon from northern and central Europe, extending to Spain, Albania and east-central Russia,
bulbilifer
Arcangeli from south-central and eastern Europe, subsp. chrysocephalus
subsp. (Reichenb.) (4)
calthifolius
(3)
&
Rouy Fouc from southern
from Greece and Crete, and subsp. ficariiformis (F.W. Schwartz)
P.D. Sell (5)
been
Europe. Plants are known to be diploid (2n=16), triploid (2n=24), or tetraploid (2n=32). Diploids have
and
referred to subspecies calthifolius and ficaria, and tetraploids to bulbilifer, chrysocephalus, ficariiformis
may hybrids and have apparently been collected
(Greilhuber 1974; Sell 1994). Triploids represent putative
&
suggested
from widely separated localities in Europe (Marchant Brighton 1974; Sell 1994). Sell (1994)
and
non-viable
that a large proportion of the pollen of triploids, as well as the tetraploid subsp. bulbilifer, is
and
few seeds are Pollen from diploids and the large-flowered tetraploids (i.e., chrysocephalus ficariiformis)
set.
Two producing
many produced subspecies are capable of
apparently viable and achenes are (Sell 1994).
is
and
and Subspecies tends to exhibit globose bulbils
bulbils in their leaf axils: bulbilifer ficariiformis. bulbilifer
subsp. produces ellipsoid bulbils (Sell 1994).
ficariiformis
no
European North American treatments either recognized subspe-
In contrast accounts, floristic
to
&
taxa in (Fernald 1950; Gleason 1952; Gleason Cronquist 1963; Whittemore 1997) or only
cific R. ficaria
&
Lambinon; Magee Ahles However, the recent
1999).
Marsden-Jones (=subsp.
variety bulbifera bulbilifer
and
discoveries of entities putatively referable to subsp. ficariiformis in North Carolina (Krings et al. 2005)
and dimensions within the
Texas (Nesom 2008; reported as subsp. but with ellipsoid bulbils flower
bulbilifer,
maybe North America
caused us question whether additional subspecies present in
range of to
ficariiformis)
and
that have not been previously recorded and if so, whether these differed in their distributions, habitats,
and
North American treatments did not emphasize subspecific recognition
rates of spread. Because prior
how many
and morpho-
were determine
subspecies are not uniformly accepted, our objectives to: (1) if
as
what such
logically recognizable entities within R. ficaria occur in the United States, (2) evaluate to extent
work and
correspond the subspecific concepts followed in Europe (based on the of Sell 1994),
entities to
analyze the distributions, habitats, and rates of spread of each entity.
(3)
from
and information were recorded from 319 herbarium specimens, requested the follow-
Distribution habitat
&
Gleason Cronquist
based on previous reports (Benson 1942; Bell 1945;
ing forty-seven herbaria literature
HNH,
CONN, GH,
1991; Whittemore 1997): A, AUA, BALT, BH, BKL, BRIT, CU, DOV, F, FLAS, GA, ILLS,
NHA,
KE, LGO, LSU, MARY, MASS, MICH, MISS, MO, MOR, MSC, MT, MU, NA, NCSC, NCU, NY, OS,
OSC, PH, POM, TENN, TEX, UNA, US, USF, USCH, VDB, VPI, WTU, WVA, Y (Appendix A). Herbarium
label data recorded for each specimen included collector name, collector number, date, habitat, and county
and Specimens lacking information were excluded from the study.
state of collection.
Collection were classified into the following nine habitat classes: (1) adjacent to a water source,
localities
and
moist
(2) disturbed areas, (3) dry woods, (4) fields, (5) horticultural, (6) lawns, (7) lowlands, (8) areas,
roadsides. A specimen was classified as adjacent to a water source if it was collected along the banks of a
(9)
stream, or pond. Moist areas were defined as moist or alluvial woods, swamp or bog areas, and other
river,
moist shade. The horticultural class was defined as being collected in a nursery or garden under cultivation.
Lowlands were defined as low or depressed areas where moisture level was not mentioned on the label.
and
lawn and sun such pastures
Fields were defined as any open grassy area not maintained as a in full as
how many within occur in the
To determine and morphologically recognizable entities R. ficaria
if
(OTU)
United each herbarium sheet was treated as an operational taxonomic unit for data capture
States,
what such correspond the subspecific
and pertinent subsequent analyses. To evaluate to extent entities to
concepts followed in Europe, each specimen was determined to subspecies following the key constructed
by
Sell (1994):
not present in leaf axils after flowering; achenes well developed
Is
present in leaf axils after flowering; achenes poorly developed _
Is
Sell (1994) recognized the difficulty in identifying Ranunculus ficaria to the subspecific level, noting that
We
specimens should be examined throughout the growing season for positive identifications. agree with
him Nesom
and others (Whittemore 1997; 2008) that identification can be challenging and recognize the
we
impact identifications have on analysis results. However, feel reasonably confident in our subspecies
own
assignments due in part to the quality of specimens, which facilitated taxon assignment, as well as our
field observations and phenological analyses. Of the 319 herbarium specimens examined 232 had at least a
month 90% 5%
and year date on the label and of these were collected mid-April through June. Another were
collected in the last week of March and the remaining 5% were collected in January through mid-March.
Based on date of collection, the majority of specimens examined in this study were collected late enough in
would
the spring that they exhibit bulbils they were genetically capable of producing them.
if
Using a digital caliper, the following morphological measurements were taken from each OTU: leaf
(1)
length from up to ten leaves, (2) leaf width from up to ten leaves, (3) petiole length from up to ten leaves
(using same leaves measured for length and width), petal length from up to ten flowers, petal width
(4) (5)
from up to ten flowers, (6) achene length of all achenes present, (7) achene width for all achenes present.
The presence or absence of bulbils was also recorded.
Quantitative and qualitative data were studied jointly and separately. Statistic analyses, including
&
ANOVAs
R
and post-hoc tests (Tukey's HSD), were carried out in the statistics package (Ihaka Gentleman
R
Foundation Computing we
1996; for Statistical 2008). Prior to multivariate analysis, tested quantita-
all
tive univariate variables using the Shapiro -Willks normality test and subsequently log transformed them
10
minimize on
to the influence of allometry the results (Dufrene 1991; Almeida-Pinheiro de Carvalho
et
al.
et al. 2004; Pimentel et al. 2007). Gower's dissimilarity coefficient for mixed data was used to quantify
OTUs OTUs
resemblances between (Gower 1971). The relationships between were subsequently explored
with both hierarchical agglomerative cluster analyses and principal coordinates analyses (PCoA) using the
complete set of characters. Three different sorting algorithms were used to help distinguish between data-
&
dependent and potential method-dependent differences in results, following Dickinson Phipps (1985) and
&
(UPGMA;
Pimentel et (2007): single linkage, complete linkage, and average linkage Sneath Sokal 1973).
al.
Quantitative characters were also analyzed separately using Principal Components Analysis (PCA).
PCA
an
is objective, correlation-based technique that allows the variance in quantitative characters to be
considered simultaneously and the subsequent visualization of dispersion patterns in a number of dimen-
&
sions that explain the greatest amount of variance (Sargent 2004; Bruneau 2007; Pimentel
et al. Joly et
A PCA
al. 2007). Kaiser-Meyer- Olkin (KMO) test was performed prior to the to assess the suitability of the
data for multivariate analysis (see also Almeida-Pinheiro de Carvalho et al. 2004; Pimentel et al. 2007).
Classification trees were employed to help identify specific morphological characters that could reliably
OTU
separate the groups corresponding to the five putative subspecies sensu Sell (1994). Classification trees
divide datasets with pre-assigned group membership into increasingly homogeneous subsets in tree-like
fashion based on the included morphological Recovery proceeds until the groups obtained are pure or
traits.
&
until a dividing threshold is achieved (Joly Bruneau 2007). For the classification tree, all morphological
characters were included and quantitative data was not log transformed prior to analysis.
The distribution of R. ficaria was mapped based on herbarium specimens and using ArcGIS 9.2 (ESRI
2004). The rate of spread for each subspecies was determined based on the number of counties each was
present in during each decade from 1860 to the present. These data were analyzed using proc mixed in SAS
journal of the Botanical Research Institute of Texas 3(1)
we
with value of 0.05 (SAS Institute 2002-2005). Note that use "spread" in a broad sense, as
a critical
3
;uspect that not new county records are the result of physical movement of propagules from estab-
all
many may
»d parental plants, but that reflect novel introduction events.
Descriptive statistics
mea-
Table provides non-transformed means and standard deviations for the seven quantitative characters
1
OTUs and summarizes
sured for each group of assignable to one of five putative subspecies sensu Sell (1994)
one-way ANOVAs and subsequent post-hoc (Tukey's HSD) on log transformed data
the results of the tests
10
showing non-transformed quantitative
Figure exhibits box-plots the distribution of
(see superscripts). 1
OTU OTUs
measurements taken by group. Significant differences in the means for each of the five groups of
were found Post-hoc
assigned to the subspecies sensu Sell (1994) for all seven characters (Table 1). tests
HSD) determine which sample means differed from which others showed that means of leaf
(Tukey's to
OTU
length and petal width of the group assignable to subsp. chrysocephalus differed significantly (p<0.05)
OTUs
OTUs
from the respective means of the groups of assignable to the four other subspecies (Table 1).
mean
assignable to the diploid subsp. calthijolius and ficaria differed significantly (p<0.05) in leaf length,
mean
mean width, and mean petiole length (Table They did not differ significantly in petal length,
leaf
1).
OTUs
mean width, mean achene length, or mean achene width. assignable to the tetraploid subsp.
petal
mean
and mis differed significantly (p<0.05) from one another in leaf length
chrysocephalus,
bulbilifer, ficariij'or
and mean petal width. Subspecies differed significantly (p<0.05) from both subsp. chrysocephalus
bulbilifer
and mean width, mean petiole length, and mean petal length. Subspecies chrysocephalus
in leaf
ficariiformis
differed significantly (p<0.05) from subsp. bulbilifer in mean achene length and width, but subsp. ficariiformis
differed neither from subsp. chrysocephalus nor bulbilifer in these characters (Table 1).
Cluster analyses
—
—
OTUs
two
In three cluster analyses average, complete, and single linkage were resolved into large
all
and sensu
divisions, these corresponding to the bulbil bearing taxa: subsp. bulbilifer subsp. ficariiformis
(1)
and
subsp.
Sell (1994) and, (2) the non-bulbil bearing taxa: subsp. calthifolius, subsp. chrysocephalus, ficaria
from
sensu Sell (1994; Fig. Within these two divisions, the topologies resulting the three different
2).
Within
from the
algorithms differed notably only for those resulting single linkage. bulbilifer/ficariiformis
OTUs
composed
division, average linkage recovered a cluster predominantly of referable to subsp. ficarii-
OTUs predominantly subsp. Both recovered clusters
formis sister to a larger cluster of referable to bulbilifer.
OTUs The complete linkage analysis recovered three clusters within
contained referable to either subspecies.
the division— one of OTUs predominantly referable to ficariiformis, nested within two
bulbilifer/ficariiformis
composed predominantly of OTUs referable to subsp. bulbilifer. Single linkage similarly recovered a cluster
OTUs
of OTUs referable to subsp. ficariiformis nested within referable to subsp. bulbilifer. In the calthifolius/
OTUs
both average and complete linkage analyses recovered a cluster of
division,
chrysocephalus/ficaria
OTUs
predominantly referable to subsp. ficaria nested within clusters of predominantly referable to subsp.
OTUs referable to subsp. ficaria did not emerge in a distinct cluster in the single linkage analy-
calthifolius.
OTUs
subsp.
but rather were interspersed throughout those referable to subsp. calthifolius. referable to
sis,
OTUs
chrysocephalus emerged interspersed in grades of referable to subsp. calthifolius and subsp. ficaria in a
sister position to the rest of the division in all three analyses.
PCA
PCoA
and
OTUs PCoA
Consistent with cluster analysis results, two non- overlapping clusters of were recovered in the
and
corresponding to the bulbil bearing taxa: subsp. bulbilifer and subsp. ficariiformis sensu Sell (1994)
(1)
and sensu
the non-bulbil bearing taxa: subsp. subsp. chrysocephalus, subsp. ficaria Sell (1994;
calthifolius,
(2)
OTUs
was by each group one
Fig. 3A). Within both of these clusters, cohesiveness exhibited of referable to
some
of the five subspecies sensu Sell (1994), although each group overlapped with another to degree.
OTUs
were
Consistent with expectations for infraspecific entities, distinct but overlapping clusters of
Leaf length 2.18 a-d (0.41; 82) .93 a (0.46; 96) 3.54 b (1.03; 6) >.42 d (0.39; 37) 2.97 c (0.49; 2) 9.79
1 1
b d d ^d
Leaf width 2.83 a (0.53; 82) 2.38 (0.54; 97) 4.33 (1.1 5; 6) (0.48; 37) 3.80 (0.53; 2) 10.56
1
(cm)
^
^
Petiole length 10.41 a (2.79; 82) 6.55 b (1.62; 97) 16.21 c (1.44; 6) a d (2.16;37) 14.75 (3.85; 12) 28.44
M
Petal length 10.23 a (1.67; 75) 1 1.1 1 a (1.91; 97) 17.43 (3.79; 6) a'c (1.95;37) 1 3.91 c-d (1.95; 11) 16.89 <0.001
w
b c d
Petal width 3.57 a (0.91; 75) 4.53 .05; 97) 7.38 (0.45; 6) 4.83 (1.09; 37) 5.75 (1.14; 23.62 <0.001
(1 1 1)
* ^
Achene length 2.95 a (0.65; 53) 3.04 a (0.65; 28) 4.33 b'c (0.69; 5) 3.27 (0.41; 6) 3.50 (0.90; 7) 5.30
* ^
Achene width 4.35 a (0.90; 53) 4.26 a .00; 28) 6.22 (0.77; 5) 12; 6) 4.70 (1.31; 7) 3.83
(1
PCA OTUs when
exhibited in the comprising all symbol coded for a priori subspecies assignments following
Among OTUs
the infraspecific concepts of Sell (1994) (Fig. 3B). these, referable to subsp. bulbilifer exhibited
the most cohesive and least diffuse cluster. In this analysis, 81% of the variation is explained by the first two
PC2
axes. PCI positively correlated most strongly with petiole length, leaf width, and leaf length, whereas
is
positively correlated most strongly with petal width and petal length (Table
is 2).
A OTUs
and
separate analysis of only referred to the two diploid taxa subsp. calthifolius ficaria, resulted
%
in two very well-defined clusters with minor overlap (Fig. 3C; Table In this analysis, 73 of variation is
3).
explained by the two axes. PCI positively correlated most strongly with petiole length, leaf length, and
first is
leaf width, whereas PC2 positively correlated most strongly with petal width and petal length (Table
is 3).
—
—
An OTUs
and
analysis of only referred to the tetraploid taxa subsp. bulbilifer, chrysocephalus, ficariijor mis
among
showed evident clustering, but with greater overlap the three a priori defined subspecies (Fig. 3D;
OTUs
Table 4). In this analysis, 76% of the variation is explained by the first two axes. defined a priori as
subsp. occupied a central coordinate space in the tetraploid analysis, flanked along the primary
ficariiformis
by and PCI most
axis subsp. chrysocephalus to the left bulbilifer to the right. is negatively correlated strongly
with petal width, leaf length, and leaf width, whereas PC2 positively correlated most strongly with petiole
is
length (Table
4).
Among
by and
the five subspecies of R. ficaria recognized Sell (1994), only the tetraploid subsp. bulbilifer
axe known to produce bulbils. A separate analyses of only OTUs with bulbils, showed two rather
ficariiformis
cohesive clusters with limited overlap corresponding to a priori assignment to these two subspecies sensu
69% by two PCI
Sell (1994; Fig. 3E; Table 5). In this analysis, of the variation is explained the first axes.
PC2
negatively correlated most strongly with petal width, whereas positively correlated most strongly
is is
A OTUs
with petal width and petal length (Table separate analysis of without bulbils (Fig. 3f; Table
5). 6),
showed three rather cohesive clusters with limited overlap, corresponding to a priori assigned subspecies.
In this analysis, 79% of the variation is explained by the first two axes. PCI is positively correlated most
PC2
strongly with petiole length, leaf width, and leaf length, whereas positively correlated most strongly
is
with petal width and petal length (Table
6).
Classification tree
showed 95% OTUs we
Classification tree analysis that or greater of the referred to subsp. calthifolius, subsp.
homogeneous
and subsp. using Sell (1994) could be placed into corresponding groups (Fig.
ficaria, bulbilifer 4).
tt[L£
;~
M
«m,
.. chrysocephaius Hear* «c or.ls
talblllft,
buib Mlthifo,iu lu ficaria ,ic ormis
h
;
—
—
—
~~
~
^socep
>„,„„.. .,«,,„,.... h, caria f,can„orml3
= :
—
—
"'™°
~:: - bulbili,er h *' """°'m s
'
Mi,hi,oMus fi ria ,ic °rmi!
h
Post et A morphometric analysis of Ranunuclus ficaria
al.,
f
s
bu ch ca
fc
ff
,
<•
;?£.•
Table 2. Character loadings combined PCA of all OTUs.
Leaf length
Leaf width
0.5562742
Petiole length
0.1484652
length
Petal
0.2944399
Leaf length 0.4393161 0.1933816 -0.5527152
Leaf width 0.4473662 0.1682932 -0.4790156
-04948622
0.7371736
Petiole length
0.1785614 0.4539478 0.2102479
Petal length
width 0.1776393 0.6952149 0.4588473
Petal
\. Character loadings for the first three principal components (PC) in the PCA of the putative tetraploid OTUs.
Leaf length -0.4145495 0.2908487 -0.5513932
Leaf width -0.4139542 0.2858958 -0.4349598
length -0.4282971 0.5777908 0.6924294
Petiole
length -0.3705260 -0.3867720 0.1113257
Petal
width -0.5797091 -0.5918072 0.1221624
Petal
Table 5. Character loadings for the first three principal components (PC) in the PCA of the bi
-0.4166727 -0.2045966 0.5826665
Leaf length
Leaf width -0.4395120 -0.2045104 0.4630089
-0.4585246 -0.6345035 -0.6193181
Petiole length
-0.3303427 0.3747980 -0.1413127
length
Petal
-0.5602160 0.6109399 -0.2063936
Table 6. Character loadings for the first three principal components (PC) in
Leaf length 0.4474063 0.1388268 -0.5661383
Leaf width 0.4522951 0.1139208 -0.4887299
04608629
Petiole length 0.6954902 -0.5497825
0.2280402 0.4510623 0.2216459
Petal length
width 0.2440242 0.6797288 0.4232127
Petal
OTUs
we and
Seventy-five percent (N=6) and 87.5% (N=7) of referred respectively to subsp. chrysocephalus
A
homogeneous
subsp. ficariiformis using Sell (1994) could be placed into corresponding groups. quarter
OTUs we
were specimens
of the (N=2) assigned to subsp. chrysocephalus in the analysis referred to subsp.
The group from
ficaria using Sell (1994). presence of bulbils separated the calthifolius/chrysocephalus/ficaria
Present
Bulbils
73 7 (87.5%)
(94.8%)
bulbilifer fi
2.5%)
1(1.3%)
calthifolius 1 (1 /?
3 (3.9%)
fh
chrysocephalus
ficaria
(96.9%) 6 (75%) chrysocephalus
ficaria
1
%) 2 (25%)
(3.1 calthifolius ficaria
OTUs
the group. Within the former, petiole length discriminated best between referable
bulbilifer/ficariiformis
to subsp. calthifolius and those referable to subsp./icaria and subsp. chrysocephalus. Petal width discriminated
best between subsp.^icaria and subsp. chrysocephalus. Petal length discriminated best between subsp. bulbilifer
we
and subsp. The shape of bulbils was not scored for the classification tree analysis as were
ficariiformis.
what
interested in seeing additional vegetative character distinguished these putative taxa.
—
Subspecies recognition. The combined results indicate the presence of five entities that can be reason-
ably referred to the subspecies accepted by Sell (1994). If one accepts subspecies as incompletely diverged
OTUs PCoA PCA
lineages, one would expect a limited amount of overlap of as seen in our and results, as
OTUs
well as incompletely sorted as seen in our cluster and classification tree analyses (Rosen et al. 2007).
ANOVA OTU
The results are also informative on this issue, particularly because the assignment of each to
a putative subspecies was based exclusively on the key by Sell (1994; see above). In this key, quantitative
measurements were used only distinguish two groups of subspecies chrysocephalus/ficariiformis and
to
(i.e.,
Qualitative characters are used in Sell's (1994) key to distinguish individual
bulbilifer/calihifolius/ficaria).
subspecies within these two groups. Thus, contributing evidence of the morphological cohesiveness of the
which found
subspecies concept of Sell (1994) is the extent to differences in quantitative characters are be-
OTU
tween subspecies pairs. Of course, had we found that our groups assigned to the subspecies sensu
all
Sell (1994) did not differ significantly in quantitative characters, would not necessarily have challenged
it
may
Sell's concepts, as the taxa truly differ only in qualitative characters. However, the finding that the
OTU
groups corresponding to the subspecies sensu Sell (1994) do in fact differ in various combinations of
the quantitative characters we examined provides some additional evidence of distinctness. Although the
means and showed
groups differed primarily in the of quantitative characters overlap
ii
sions, one accepts a subspecies as an incompletely diverged lineage, overlap in charac
if