Int. J. Plant Sci. 164(6):899–905. 2003.
䉷 2003 by The University of Chicago. All rights reserved.
1058-5893/2003/16406-0008$15.00
NO EVIDENCE OF SEX-DIFFERENTIAL POLLEN LIMITATION AT THE FLOWER LEVEL IN THE
GYNODIOECIOUS GYPSOPHILA REPENS INFECTED BY MICROBOTRYUM VIOLACEUM
Manuela López-Villavicencio,1 Carine L. Collin, and Jacqui A. Shykoff
Laboratoire d’Ecologie, Systématique et Evolution, Université de Paris-Sud (XI), Bâtiment 360, F-91405 Orsay Cedex, France
Because in gynodioecious species females are expected to be more pollen limited than hermaphrodites, we
tested for sex-differential pollen limitation at the flower level, comparing fruit production and fecundity between
open-pollinated and hand-pollinated flowers of Gypsophila repens (Caryophyllaceae) over two years. No sex
difference in pollen limitation was found. Hand-pollination increased fruit and seed production in the first
year of the study but not in the second, and it actually decreased fruit set in the latter year, although overall
fecundity was higher in that year. Ovule number differed between years, with perfect flowers containing more
ovules in the second year. Gypsophila repens harbors the anther smut fungus Microbotryum violaceum, which
infects and sterilizes many species of Caryophyllaceae. In G. repens, however, some infected plants are capable
of seed production, although their fecundity is far lower than that of healthy plants. Seed production was not
pollen limited even with the high percentage of infected individuals that could reduce the quantity of pollen
available at the population level.
Keywords: anther smut disease, Caryophyllaceae, gynomonoecy-gynodioecy, Gypsophila repens, infection,
reproductive success.
Introduction
L. (Caryophyllaceae) in two different years. Pollen limitation
at the flower level is a controversial issue. Differential seed
production by flowers to which pollen has been added does
not necessarily demonstrate that inadequate pollen is available
from natural pollination; it could also result from resources
being preferentially allocated to flowers receiving more pollen
(Zimmerman and Pyke 1988). Although we cannot rule out
this possibility, we have some evidence of no reallocation of
resources toward pollinated flowers in this species. We performed hand-pollinations on different proportions of flowers
on plants in the greenhouse; plants of each sex experienced
various pollination treatments that included pollination of all
the flowers produced or just a small percentage. So far, our
results showed no differences between treatments (M. LópezVillavicencio, unpublished data). In addition, G. repens flowers over several months and can produce thousands of flowers
per plant, rendering pollen addition experiments at the plant
level entirely unfeasible.
This species has three functional sexes: females bearing only
pistillate flowers, hermaphrodites with uniquely perfect flowers, and gynomonoecious individuals bearing a mixture of pistillate and perfect flowers. Pistillate flowers on female compared with gynomonoecious hermaphrodites could be
expected to be more pollen limited through a lack of pollen
production within the same plant if geitonogamous pollen flow
from perfect to pistillate flowers assures pollination. In addition, G. repens can be infected by the anther smut fungus
Microbotryum violaceum (Pers.) Deml. & Oberw. (pUstilago
violacea Pers.). This fungus causes the production of sporefilled anthers that bear no pollen in both pistillate and perfect
flowers. Because infection prevents pollen production, high
disease prevalence can contribute to pollen limitation for
healthy flowers in natural populations (Alexander 1987). In
Gynodioecious populations contain hermaphrodites and females that produce no pollen. Reproduction by females in selfcompatible species might be more limited by pollen availability
than that of their hermaphrodite counterparts (Lloyd 1974;
Maurice and Fleming 1995). Pistillate flowers in gynodioecious
species are usually smaller than perfect ones (Delph 1996),
rendering them less attractive to pollinators and potentially
more likely to be pollen limited when pollen production decreases in the population. For species such as Silene vulgaris,
female reproductive success decreases when the frequency of
females in the population rises (McCauley and Brock 1998).
In Glechoma hederacea, fruit and seed set in females is reduced
with increased distance from hermaphrodite pollen donors
(Widén and Widén 1990). Furthermore, pistillate flowers in
gynodioecious species such as Stellaria longipes, Iris douglasiana, and Geranium richardsonii receive less pollen than do
perfect ones in natural populations (Philipp 1980; Uno 1982;
Williams et al. 2000). Thus, pollen availability may influence
the evolution of floral sexual size dimorphism in dioecious
(Vamosi and Otto 2002) and gynodioecious populations (Ashman and Diefenderfer 2001).
Some studies have already examined pollen limitation of
females and hermaphrodites in gynodioecious populations
(Widén 1992; Fleming et al. 1994; Graff 1999; Ashman and
Diefenderfer 2001). Here, we tested for differential pollen limitation at the flower level for the different sex phenotypes in
the gynomonoecious-gynodioecious species Gypsophila repens
1
Author for correspondence; telephone 33 1 69 15 56 64; fax 33
1 69 15 73 53; e-mail [email protected].
Manuscript received January 2003; revised manuscript received May 2003.
899
INTERNATIONAL JOURNAL OF PLANT SCIENCES
900
Table 1
Repeated-Measures ANOVA on Fruit Set in 1998 and 2001 by the
Seven Flower Categories of Gypsophila repens under
the Two Pollination Treatments
Source
dfa
Year
1, 245
Category
6, 234
Category # year
6, 234
Treatment
1, 616
Treatment # year
1, 616
Treatment # category
6, 616
Year # treatment # category
6, 616
Plant (year, category)
194, 616
Error
616
Mean
squarea
3.46
8.54
1.16
0.26
2.03
0.24
0.11
0.31
0.121
F
P
13.27 0.0003
31.85 !0.0001
4.34 0.0004
2.11 0.15
16.69 !0.0001
1.94 0.07
0.87 0.51
2.52 !0.0001
Note. The two pollination treatments are open- and handpollination. R2 p 0.64. Individual plant was a random blocking factor.
a
Calculated by Satterthwaite approximations in JMP (SAS Institute
1997).
many host species, anther smut disease also leads to complete
female sterility (Baker 1947), although not in G. repens. In
the studied population of G. repens, infected plants were either
totally or partially infected, wherein partially infected plants
produced a mixture of healthy and diseased flowers. Variation
in the effect of this fungus on female structures was also observed. Whereas most infected plants bore flowers with greatly
reduced female organs, as observed in other infected species
(Shykoff et al. 1997), some had well-formed, apparently functional stigmas. In other studied systems, diseased plants are
less attractive to pollinators and receive fewer visits than
healthy ones (Jennersten 1988; Shykoff and Bucheli 1995).
Therefore, we investigated pollen limitation at the flower level
in diseased flowers.
Because environmental conditions vary over time, pollinator
activity and resource availability may change from one reproductive season to the next or even over the season (Dudash
and Fenster 1997; Parra-Tabla et al. 1998; Baker et al. 2000;
Larson and Barrett 2000). In particular, in gynodioecious species, pollen availability will vary with operational sex ratio
(McCauley and Brock 1998; Graff 1999) but also with resource variation if hermaphrodites alter allocation to male reproductive effort as a function of resource availability (Eckhart
and Chapin 1997; Poot 1997).
We examined the following questions: (1) Does pollen availability limit reproduction at the flower level? (2) Does pollen
limitation differ for healthy perfect flowers compared with
healthy pistillate flowers and/or between years? (3) Can infected flowers produce fruits and seeds, and are they pollen
limited? (4) Do healthy flowers on partially diseased plants
have reproductive characteristics of healthy or diseased flowers? Last, we discuss the implications of our findings for the
evolution of sexual dimorphism in gynodioecious species.
Material and Methods
Gypsophila repens is a subalpine-alpine perennial geophyte
distributed in the mountains of south and central Europe. It
typically occurs in sunny and, periodically, in dry places in
rocky and grassy habitats (Alegro et al. 2000). The population
we studied, containing several thousand plants, grows on an
exposed rock slope with southeast exposure in the Italian Alps
(Grosio, lat. 46⬚17⬘24⬙N, long. 10⬚15⬘11⬙E). Flowering begins
in early June and continues until late October. Individuals can
be large, some bearing hundreds of branches and several hundred flowers at a time. This population is infected with Microbotryum violaceum, and despite intensive searching, it is
the only infected population we have found.
We followed the fate of ca. 800 marked individuals over
two nonconsecutive years, noting their size, sex, and infection
status. In 1998, only 45.9% of these individuals flowered;
20.4% were infected, 4.4% were partially infected, and 75.2%
were healthy, including 9.9% females, 62.4% pure hermaphrodites, and 2.9% gynomonoecious hermaphrodites. In 2001,
63.8% of the individuals flowered; 18.7% were infected,
10.2% were partially infected, and 71.1% were healthy, of
which 8.3% were females, 54.5% were pure hermaphrodites,
and 8.3% were gynomonoecious hermaphrodites. Sexual
status of marked individuals that flowered both years remained
constant over the time (J. Shykoff and M. López-Villavicencio,
personal observation).
Perfect flowers are protandrous and self-compatible (M.
López-Villavicencio, personal observation). Pistillate flowers
were significantly smaller (mean petal length in millimeters Ⳳ
SE: 2.61 Ⳳ 0.21, n p 10) than diseased flowers (3.47 Ⳳ 0.09,
n p 54), which were significantly smaller than perfect flowers
(4.19 Ⳳ 0.07, n p 82; F(2, 143) p 36.37, P ! 0.0001, with a
posteriori Tukey’s multiple comparison). Natural pollinators
of G. repens include syrphid flies and small solitary bees. Pollinators visit many flowers on the same plant before moving
to the next plant. Pollen and spores can be transported to
plants located several meters away from the original plant (M.
López-Villavicencio, personal observation).
In July 1998 and July 2001, plants that bore at least four
open flowers were chosen for the pollination experiment.
Table 2
Repeated-Measures ANOVA on Seed Number per Fruit in 1998
and 2001 by the Six Flower Categories of Gypsophila repens
under the Two Pollination Treatments
Source
Year
Category
Category # year
Treatment
Treatment # year
Treatment # category
Year # treatment # category
Plant (year, category)
Error
dfa
Mean
squarea
1, 159
5, 164
5, 164
1, 194
1, 194
5, 194
5, 194
194, 616
194
128.69
140.53
15.83
3.48
153.76
3.18
17.01
0.31
18.15
F
P
2.98
3.33
0.38
0.19
8.47
0.18
0.94
2.52
0.09
0.007
0.87
0.66
0.004
0.97
0.46
!0.0001
Note. The two pollination treatments are open- and handpollination. R2 p 0.75. Diseased flowers on partially diseased plants
were excluded because too few data were available. Individual plant
was a random blocking factor.
a
Calculated by Satterthwaite approximations in JMP (SAS Institute
1997).
LÓPEZ-VILLAVICENCIO ET AL.—POLLEN LIMITATION IN GYPSOPHILA REPENS
901
Fig. 1 Percentage Ⳳ binomial error of flowers of Gypsophila repens setting fruits in 1998 and 2001 under two pollination treatments, i.e.,
open-pollination (white and black bars) versus hand-pollination (hatched bars). Seven flower categories are considered: pistillate flowers on
female (FF) and gynomonoecious (FM) plants, perfect flowers on gynomonoecious (HM) and hermaphrodite (HH) plants, healthy flowers on
plants partially diseased with Microbotryum violaceum (HD), diseased flowers on partially diseased (DD), and totally diseased plants (Diseased).
Numbers below the bars denote the number of flowers per treatment per year. In 1998, pollen limited fruit production only for perfect flowers
(t-test allowing unequal variances: t p 3.12, P p 0.003); see table 1.
Plants belonged to several types, varying in sex phenotype and
infection status. In 1998, the plants used included 60 healthy
individuals (17 females, 29 hermaphrodites, and 14 gynomonoecious), seven partially infected, and 32 totally infected
plants, with overrepresentation of those with well-developed,
apparently functional stigmas. In 2001, the selected plants included 50 healthy individuals (15 females, 23 hermaphrodites,
12 gynomonoecious), nine partially infected, and 18 completely infected plants chosen at random.
Depending on availability, two or three flowers were handpollinated, and the same number was left unmanipulated for
open-pollination (hereafter referred to as “controls”). Pollinations were performed by brushing two or three anthers,
presenting ripe pollen across the stigmatic surface. Pollen is
bluish violet, and full anthers are readily distinguished from
empty, pale violet anther sacs. The anthers were taken from
at least two different individuals of the same population but
located several meters away from the pollen recipients. Pollinated and control flowers were marked with a small piece of
labeled tape on the pedicel. In gynomonoecious plants, two
open flowers of each sex were pollinated and two of each sex
were designated as controls.
To estimate the effect of hand-pollination on fecundity, fruits
were collected as they matured (ca. 15 d after the pollination),
and the seeds were counted. Because differences in seed number
per fruit could result from disparity in ovule number, in 1998
and 2001, a sample of pistillate and perfect flowers was collected into vials of 95% ethanol, and the ovules were counted
under a dissecting microscope. In 2001, we also counted ovules
from infected flowers.
All statistical analyses were performed with JMP, version
4.0 (SAS Institute 1997). To investigate pollen limitation for
fruit set and seed number per fruit, we calculated the difference
between the means of pollinated and control flowers for each
flower type within each plant. This difference was then subjected to two different two-way ANOVAs that tested the effects
of year and flower category. Some pseudoreplication remains
in the analyses because gynomonoecious individuals and partially diseased plants were represented by more than one difference value, which were treated as independent observations.
To investigate the reproductive characteristics of the different types of flowers in our population, we analyzed fruit set
and seed number per fruit data using repeated-measures threeway ANOVAs, treating the individual plant as a random blocking factor. Diseased flowers on partially diseased individuals
set too few seeds to be included in the analysis of seed number
per fruit.
Residuals from the analyses of seed number per fruit did
not deviate from normality, but those from the fruit set analyses were slightly bimodal, and no data transformation alleviated this problem. We present these data nonetheless, noting
that the deviations from normality were minor, but sample
sizes were large, providing statistical power to detect the deviations (residuals from the analysis presented in table 1:
Shapiro-Wilks’s W p 0.944, n p 196, P ! 0.0001; residuals
902
INTERNATIONAL JOURNAL OF PLANT SCIENCES
Fig. 2 Mean Ⳳ SE number of seeds produced by fruits of Gypsophila repens in 1998 and 2001 under the two pollination treatments, i.e.,
open-pollination (white and black bars) versus hand-pollination (hatched bars). Seven flower categories are considered: pistillate flowers on
female (FF) and gynomonoecious (FM) plants, perfect flowers on gynomonoecious (HM) and hermaphrodite (HH) plants, healthy flowers on
plants partially diseased with Microbotryum violaceum (HD), diseased flowers on partially diseased (DD), and totally diseased plants (Diseased).
Numbers below the bars denote the number of flowers per treatment per year. In 1998, pollen limited seed production for perfect flowers
(contrasts t-test: t p 2.19, P p 0.030); see table 2.
from the analysis presented in table 2: Shapiro-Wilks’s W p
0.937, n p 838, P ! 0.0001).
Results
Fruit Set
The difference in fruiting success between pollinated and
control flowers was small but positive (mean difference Ⳳ SE:
0.059 Ⳳ 0.028, n p 133, t p 2.15, P p 0.03), indicating
some overall pollen limitation of fruit production at the flower
level. The degree of pollen limitation for fruit set differed between years (F(1, 616) p 16.69, P ! 0.0001; treatment # year interaction in table 1), with the pollen addition treatments increasing fruit set in 1998 and decreasing it in 2001, but this
did not differ among the seven flower categories (pistillate
flowers on female or gynomonoecious individuals, perfect
flowers on hermaphrodite or gynomonoecious individuals,
healthy flowers—pistillate and perfect combined—on partially
diseased plants, or diseased flowers on totally or partially diseased plants; F(6, 616) p 0.87, P p 0.51).
Fruit set varied across plant individuals and between years,
with globally higher fruiting success in 1998 than in 2001.
This was mainly caused by the greater success of the diseased
flowers in 1998, which was revealed by a significant
category # year interaction (table 1). Over the 2 yr, only
19.6% of diseased flowers set fruit, whereas 72.9% of healthy
flowers set fruit. In 2001, only 5.1% of the flowers on randomly chosen infected plants produced fruits compared with
32.8% of those chosen for their apparently functional stigmas
in 1998 (fig. 1). The flower categories differed significantly for
successful fruit production (table 1), with diseased flowers setting fruits, but differed significantly more rarely than healthy
ones, regardless of whether they were borne on totally or partially diseased plants. Healthy flowers on diseased plants had
similar fruit set to healthy flowers on healthy plants (fig. 1).
Seed Number per Fruit
The difference in seed number per fruit between pollinated
and control flowers across all categories did not differ significantly from 0 (mean difference Ⳳ SE: 0.134 Ⳳ 0.686, n p
91, t p 0.195, P p 0.84), indicating an absence of pollen limitation for seed production of fruits that reached maturity. The
degree of pollen limitation for seed number per fruit differed
between years (F(1, 194) p 8.47, P p 0.004; treatment # year interaction in table 2), with the pollen addition treatment increasing seed production per fruit in 1998 and decreasing it
in 2001, but it did not differ significantly among the seven
flower categories (F(5, 194) p 0.18, P p 0.97).
Seed number per fruit differed among individual plants and
across the six flower categories investigated by being lower for
diseased than for healthy flowers. Seeds per fruit differed almost significantly between years, with higher seed production
per fruit in 2001 than in 1998 (table 2; fig. 2).
LÓPEZ-VILLAVICENCIO ET AL.—POLLEN LIMITATION IN GYPSOPHILA REPENS
Table 3
Number of Ovules Found in Pistillate, Perfect, and Infected
Flowers of Gypsophila repens in 1998 and 2001
Flower category
Pistillate
Perfect
Infected
1998
mean Ⳳ SE
n
25.73 Ⳳ 1.84
23.19 Ⳳ 0.87
15
21
2001
mean Ⳳ SE
n
25.74 Ⳳ 0.98
29.77 Ⳳ 0.77
24.84 Ⳳ 0.85
27
52
25
Ovule Production
Total ovule production was significantly higher in 2001 than
in 1998 largely because of an increase in ovule production in
perfect flowers in 2001 (table 3), which was revealed by the
significant category # year interaction (table 4). Ovule number
of infected flowers was similar to that of pistillate flowers (table
3).
Discussion
Pistillate flowers are expected to be more pollen limited than
perfect ones (Widén 1992; Graff 1999). Fecundity of both
females and hermaphrodites declines with increasing female
frequency in other gynodioecious species, presumably from
low pollen availability (McCauley and Brock 1998; Graff
1999), and pollen limitation was found in populations of Glechoma hederacea with at least 50% of female plants (Widén
and Widén 1990; Widén 1992). However, the frequency of
female plants at which pollen limitation should become evident
remains unclear, and a meta-analysis on reproductive differences between female and hermaphrodite plants in gynodioecious species found no significant difference in pollen limitation between the sexes (Shykoff et al. 2003). Here, we found
no evidence for sex-differential pollen limitation at the flower
level for either fruit or seed production in Gypsophila repens
over 2 yr of pollen addition experiments. In our population
in both years, approximately one-third of flowering individuals
bore no pollen because they were either female or diseased.
Furthermore, not only do diseased plants produce no pollen,
but they can serve as pollen traps (Alexander 1987). Spores
of Microbotryum violaceum on stigmas of healthy plants can
impede pollen germination, exacerbating effects of low pollen
availability (Marr 1998).
This combination of low population pollen production and
diseased flowers, both intercepting pollen and contaminating
healthy stigmas, resulted in pollen limitation in 1998. It is
surprising that perfect flowers were pollen limited for fruit and
seed production (figs. 1, 2) in this year. It is unclear whether
this resulted from pollen interference by fungal spores or from
poor pollinator service. To separate these factors, pollinator
observations and detailed investigations of pollen and spore
dynamics in patches or in populations of varying sex ratio and
pollen availability are called for (Graff 1999).
The effect of the pollen addition treatment differed between
years, increasing fruit and seeds per fruit in 1998 but decreasing fruit set in 2001 (fig. 1). Different people conducted the
experiments in the two years; although we followed the same
general protocol, individual differences in handling inevitably
existed. Physical stimulation and pollinator visitation influence
903
both floral longevity (Nyman 1993; Kaltz and Shykoff 2001)
and potential fruiting success. Furthermore, selective shedding
of flowers that receive many pollinator visits or much physical
stimulation early in the season could be adaptive in G. repens,
both for reducing disease risk (Shykoff et al. 1996; Kaltz and
Shykoff 2001) and for stimulating subsequent flower production. Gypsophila repens produces a large indeterminate number of flowers. In such species, flower production is often limited by successful fruit production, with plants producing few
fruits flowering longer (Carroll and Delph 1996). Plants that,
early in the flowering season, shed flowers receiving much
physical stimulation (i.e., much pollinator visitation), thereby
stimulating further flower production, may enhance their fitness via male function because both early-shed and late-shed
flowers export pollen. We intend to test the relationship between pollination intensity, fruit production, and subsequent
flower production in greenhouse experiments.
It is interesting that a difference in ovule number across years
was expressed in perfect flowers only. The reproductive characters of perfect flowers are expected to be more plastic than
in females if their reproductive strategy varies between pure
male and hermaphrodite (Lloyd 1975). However, perfect flowers are expected on average to have lower female function than
pure females because some female advantage is required to
explain female persistence in natural populations containing
hermaphrodites (Couvet et al. 1990). We found that perfect
flowers contained more ovules than pistillate ones in 2001, a
rather surprising result. Pistillate flowers are smaller than perfect ones, and flower size, pollen, and ovule production are
all intercorrelated through genetic correlations (Ashman 1999;
Mazer and Dawson 2001) and structural constraints.
The absence of sex-differential pollen limitation indicates
that seed production in both types of flowers may be limited
by resources rather than pollen (Zimmerman and Pyke 1988).
We suggest that under high resource availability large perfect
flowers can increase their ovule production, whereas small pistillate flowers cannot because of structural space limitations.
Nonetheless, female plants may have other avenues for increasing fecundity, such as longer life span or enhanced flower
production (Shykoff et al. 2003).
Some G. repens infected with the anther smut fungus M.
violaceum were able to produce fruits and viable seeds that
germinated and gave rise to healthy plants. This is the first
report we know of this phenomenon, although partial sterilization is predicted to evolve in strongly spatially structured
populations of hosts and sterilizing pathogens (O’Keefe and
Antonovics 2002). Although M. violaceum did not completely
sterilize its infected hosts, it strongly reduced fecundity (figs.
Table 4
ANOVA of the Number of Ovules for Pistillate and Perfect Flowers
Source
Year
Flower category
Flower category # year
Error
Note. R2 p 0.19.
df
Mean
square
1
1
1
111
254.3
12.9
253.2
29.3
F
P
8.690
0.442
8.651
0.004
0.508
0.004
904
INTERNATIONAL JOURNAL OF PLANT SCIENCES
1, 2) but without influencing ovule number (table 3). Microbotryum violaceum sterilizes female function of Silene latifolia
hosts by sporulating within the ovules (Batcho et al. 1979),
but fungal activity in infected female tissues of G. repens remains to be elucidated. Indeed, some completely diseased individuals were sterilized by the fungus, while others maintained functional stigmas even though their anthers produced
fungal spores in the place of pollen. We are currently testing
whether these different phenotypes of diseased plants depend
on particular fungal strains or variation in plant defense.
Pollen limitation can influence the evolution of floral display.
Fragaria virginiana females are pollen limited when hermaphrodites are rare, and pollinator preference for large-petaled
pistillate flowers will influence selection on floral sexual size
dimorphism (Ashman and Diefenderfer 2001). Indeed, pistillate flowers are almost always smaller than perfect flowers in
gynodioecious species. The degree of sexual size dimorphism
appears to depend on the potential for pollen limitation, with
species having many-seeded fruits that can more easily suffer
from pollen limitation showing less reduction in pistillate
flower size (Shikoff et al. 2003). Parasites and herbivores can
change the population sex ratio if they attack one sex preferentially (Ashman 2002). Microbotryum violaceum is transmitted by pollinators; therefore, a plant’s attractiveness to pollinators places it at a higher risk for infection (Shykoff et al.
1997). For G. repens, it would be interesting to know whether
the sexes vary in vulnerability. Different susceptibility to postcontamination infection is known for dioecious species because
female plants retain physiological connections to contaminated
flowers longer than hermaphrodites (Kaltz and Shykoff 2001),
but such a mechanism is unlikely for pistillates compared with
perfect flowers where both types of flowers mature fruits.
Nonetheless, different susceptibility of the two sexes would
have consequences for the population’s sex ratio and possibly
evolution of sexual size dimorphism.
We found temporal variation in the degree of pollen limitation for both fruit and seeds per fruit in the gynomonoeciousgynodioecious G. repens, but there was no evidence for sexdifferential pollen limitation at the level of individual flowers.
Although local pollen production would be expected to reduce
pollen limitation for hermaphrodite plants contrasted with female ones, this hypothesis requires further comparisons between populations with different sex ratios and infection
frequencies.
Acknowledgments
We thank two anonymous reviewers for their judicious comments and suggestions. We also thank Claire Pergrale for hard
work in the field and in the lab. The Park of the Rupestrian
Engravings of Grosio (Italy) kindly permitted us to work on
their site. Heidi Hauffe and Ezio Olandi gave logistic and
moral support in the field. M. López-Villavicencio was supported by a fellowship from the Consejo Nacional de Ciencia
y Tecnologı́a and C. L. Collin by the Ministère de l’Education
National de la Recherche et Technologie during this work.
Literature Cited
Alegro A, M Biljacovic, A Ostojic, M Calic 2000 Gypsophila repens
L. (Caryophyllaceae), an overlooked species in the flora of Croatia.
Acta Bot Croat 59:331–336.
Alexander H 1987 Pollination limitation in a population of Silene
alba infected by the anther-smut fungus, Ustilago violacea. J Ecol
75:771–780.
Ashman T-L 1999 Determinants of sex allocation in a gynodioecious
wild strawberry: implications for the evolution of dioecy and sexual
dimorphism. J Evol Biol 12:648–661.
——— 2002 The role of herbivores in the evolution of separate sexes
from hermaphroditism. Ecology 83:1175–1184.
Ashman T-L, C Diefenderfer 2001 Sex ratio represents a unique context for selection on attractive traits: consequences for the evolution
of sexual dimorphism. Am Nat 157:334–347.
Baker A, SCH Barrett, JD Thompson 2000 Variation of pollen limitation in the early flowering Mediterranean geophyte Narcissus assoanus (Amaryllidaceae). Oecologia 124:529–535.
Baker HG 1947 Infection of species of Melandrium by Ustilago violacea (Pers.) Fuckel and the transmission of the resultant disease.
Ann Bot 11:333–348.
Batcho M, JC Audran, C Zambettakis 1979 Sur quelques données de
l’évolution de l’Ustilago violacea (Pers.) Rouss. dans les ovules du
Silene dioica (L.) Cl. Rev Cytol Biol Veg Bot 2:329–346.
Carroll SB, LF Delph 1996 The effects of gender and plant architecture on allocation to flowers in dioecious Silene latifolia (Caryophyllaceae). Int J Plant Sci 157:493–500.
Couvet D, A Atlan, E Belhassen, C Gliddon, P-H Gouyon, F Kjellberg 1990 Co-evolution between two symbionts: the case of cytoplasmic male-sterility in higher plants. Pages 225–249 in D Futuyma, J Antonovics, eds. Oxford surveys in evolutionary biology.
Vol 7. Oxford University Press, Oxford.
Delph LF 1996 Flower size dimorphism in plants with unisexual flowers. Pages 217–237 in DG Lloyd, SCH Barrett, eds. Floral biology:
studies on floral evolution in animal-pollinated plants. Chapman &
Hall, New York.
Dudash MR, CB Fenster 1997 Multiyear study of pollen limitation
and cost of reproduction in the iteroparous Silene virginica. Ecology
78:484–493.
Eckhart VM, FS Chapin III 1997 Nutrient sensitivity of the cost of
male function in gynodioecious Phacelia linearis (Hydrophyllaceae).
Am J Bot 84:1092–1098.
Fleming TH, S Maurice, S Buchmann, M Tuttle 1994 Reproductive
biology and relative male and female fitness in a trioecious cactus
Pachisereus pringlei (Cactaceae). Am J Bot 81:858–867.
Graff A 1999 Population sex structure and reproductive fitness in
gynodioecious Sidalcea malviflora malviflora (Malvaceae). Evolution 53:1714–1722.
Jennersten O 1988 Insect dispersal of fungal disease: effects of Ustilago infection on pollinator attraction in Viscaria vulgaris. Oikos
51:163–170.
Kaltz O, JA Shykoff 2001 Male and female Silene latifolia plants
differ in per-contact risk of infection by a sexually transmitted disease. J Ecol 89:99–109.
Larson BMH, SCH Barrett 2000 A comparative analysis of pollen
limitation in flowering plants. Biol J Linn Soc 69:503–520.
Lloyd DG 1974 Theoretical sex ratios of dioecious and gynodioecious
angiosperms. Heredity 32:11–34.
——— 1975 The maintenance of gynodioecy and androdioecy in angiosperms. Genetica 45:325–339.
Marr D 1998 The effect of Microbotryum violaceum spores on pollen
germination in Silene acaulis. Int J Plant Sci 159:221–227.
Maurice S, TH Fleming 1995 The effect of pollen limitation on plant
LÓPEZ-VILLAVICENCIO ET AL.—POLLEN LIMITATION IN GYPSOPHILA REPENS
reproductive systems and the maintenance of sexual polymorphisms.
Oikos 74:55–60.
Mazer SJ, KA Dawson 2001 Size-dependent sex allocation within
flowers of the annual herb Clarkia unguiculata (Onagraceae): ontogenetic and among-plant variation. Am J Bot 88:819–831.
McCauley DE, MT Brock 1998 Frequency-dependent fitness in Silene
vulgaris, a gynodioecious plant. Evolution 52:30–36.
Nyman Y 1993 The pollen-collecting hairs of Campanula (Campanulaceae). II. Function and adaptive significance in relation to pollination. Am J Bot 80:1437–1443.
O’Keefe KJ, J Antonovics 2002 Playing by different rules: the evolution of virulence in sterilizing pathogens. Am Nat 159:597–605.
Parra-Tabla V, CF Vargas, LE Eguiarte 1998 Is Echeveria gibbiflora
(Crassulaceae) fecundity limited by pollen availability? an experimental study. Funct Ecol 12:591–595.
Philipp M 1980 Reproductive biology of Stellaria longipes Goldie as
revealed by a cultivated experiment. New Phytol 85:557–569.
Poot P 1997 Reproductive allocation and resource compensation in
male-sterile and hermaphroditic plants of Plantago lanceolata (Plantaginaceae). Am J Bot 84:1256–1265.
SAS Institute 1997 JMP, version 4.0. SAS Institute, Cary, N.C.
Shykoff JA, E Bucheli 1995 Pollinator visitation patterns, floral rewards and the probability of transmission of Microbotryum violaceum, a venereal disease of plants. J Ecol 83:189–198.
905
Shykoff JA, E Bucheli, O Kaltz 1996 Flower lifespan and disease risk.
Nature 379:779.
——— 1997 Anther smut disease in Dianthus silvester (Caryophyllaceae): natural selection on floral traits. Evolution 51:383–392.
Shykoff JA, S-O Kolokotronis, CL Collin, M López-Villavicencio 2003 Effects of male sterility on reproductive traits in gynodioecious plants: a meta-analysis. Oecologia 135:1–9.
Uno GE 1982 Comparative reproductive biology of hermaphroditic
and male-sterile Iris douglasiana Herb. (Iridaceae). Am J Bot 69:
818–823.
Vamosi JC, SP Otto 2002 When looks can kill: the evolution of sexually dimorphic floral display and the extinction of dioecious plants.
Proc R Soc Lond B Biol Sci 269:1187–1194.
Widén B, M Widén 1990 Pollen limitation and distance-dependent
fecundity in females of the clonal gynodioecious herb Glechoma
hederacea (Lamiaceae). Oecologia 83:191–196.
Widén M 1992 Sexual reproduction in a clonal, gynodioecious herb
Glechoma hederacea. Oikos 63:430–438.
Williams CF, MA Kuchenreuther, A Drew 2000 Floral dimorphism,
pollination and self-fertilization in gynodioecious Geranium richardsonii (Geraniaceae). Am J Bot 87:661–669.
Zimmerman M, GH Pyke 1988 Reproduction in Polemonium: assessing the factors limiting seed set. Am Nat 131:723–738.