Interrelationships Among Inbreeding Depression, Plasticity in the Self-incompatibility System, and the Breeding System of Campanula rapunculoides L. (Campanulaceae)

Annals of Botany 85 (Supplement A): 211±219, 2000 doi:10.1006/anbo.1999.1033, available online at http://www.idealibrary.com on

Interrelationships Among Inbreeding Depression, Plasticity in the Self-incompatibility System, and the Breeding System of Campanula rapunculoides L. (Campanulaceae) A N D R E W G . S T E P H E N S O N * , SA R A V. G O O D and D O N N A W. VO G L E R Department of Biology, The Pennsylvania State University, University Park, PA 16802, USA Received: 21 July 1999 Returned for revision: 9 September 1999 Accepted: 20 October 1999 The evolution of breeding systems in plants is often viewed as a balance between the adverse consequences of sel®ng (inbreeding depression and the loss of opportunities to sire seeds on conspeci®cs) and the bene®ts of sel®ng (a genetic transmission advantage and reproductive assurance when cross pollen limits seed production). In this paper we examine the genetic and environmental causes of variation in the expression of self-incompatibility (SI) in Campanula rapunculoides and explore the consequences of this variation on the breeding system. Campanula rapunculoides has an S-RNase based SI system similar to that described in the Solanaceae. However, our studies of plants from two natural populations have revealed that the ¯owers of most individuals are self-incompatible when they ®rst open but become more self fertile as the ¯owers age. Moreover, when both cross and self pollen are deposited onto the stigmas of older ¯owers, the cross pollen tubes grow faster and sire a disproportionate number of the seeds. In short, self-fertilization occurs only after most opportunities for outcrossing have occurred. We also found that there is signi®cant heritable genetic variation in the population for the strength of SI in young and old ¯owers and for the amount of breakdown in SI indicating that natural selection could operate on the strength of SI and its breakdown. In a multigenerational study, we used controlled crosses to create families of plants with a range of inbreeding coecients (0, 0.25, 0.5 and 0.75). We found that ®tness declined signi®cantly over the range of inbreeding coecients and that the decline in ®tness was less for families derived from weak SI phenotypes. Consequently, it is only advantageous for C. rapunculoides to produce selfed seed when seed production is limited by the availability of cross pollen. Because of plasticity in the SI system, C. rapunculoides has a breeding system that combines the best of both worlds. # 2000 Annals of Botany Company Key words: Campanula rapunculoides, self-incompatibility, pseudo self-compatibility, heritability, S-RNase, gametophytic self-incompatibility, breeding system, mating system, sel®ng, outcrossing, inbreeding depression.

I N T RO D U C T I O N Since the publication of Darwin's (1876) classic book, The e€ects of cross- and self-fertilisation in the vegetable kingdom, inbreeding depression (the reduction in ®tness of inbred progeny relative to outcrossed progeny) has been invoked as a major force in the evolution of mechanisms that promote outcrossing. On the other hand, natural historians have long noted that sel®ng is advantageous if it provides reproductive assurance when pollinators are scarce or unreliable (e.g. Baker, 1955; Stebbins, 1957; Schoen, Morgan and Bataillon, 1996), and evolutionary biologists have long noted that there is an inherent transmission advantage to sel®ng because a plant donates two haploid sets of chromosomes to each selfed seed and only one set to each outcrossed seed (e.g. Fisher, 1941). Consequently, genetic modi®ers that promote self-fertilization without reducing cross-fertilization through the male ( pollen) function should increase in the absence of inbreeding depression (e.g. Nagylaki, 1976; Charlesworth, 1980b; Lande and Schemske, 1985; Campbell, 1986). Sel®ng may also be advantageous if it maintains adaptive combinations of alleles at di€erent loci (e.g. Stebbins, 1957; Solbrig, 1976), and if it costs less, in terms of energy and other * Fax ‡1-814-865-9131, e-mail [email protected] or [email protected] or [email protected], respectively.

0305-7364/00/0A0211+09 $35.00/00

resources, to produce selfed seed (e.g. fewer resources are expended to attract and reward pollinators) (Schemske, 1978; Waller, 1979; Schoen and Lloyd, 1984). Recent theoretical advances have demonstrated that the genetic and environmental advantages of sel®ng are not only counterbalanced by inbreeding depression (e.g. Maynard Smith, 1977; Charlesworth, 1980a) but also by the adverse e€ects of sel®ng on pollen dissemination to conspeci®cs ( pollen discounting: e.g. Nagylaki, 1976; Holsinger, Feldman and Christiansen, 1984) and by a decrease in outcross seed production that may accompany an increase in sel®ng (seed discounting: Lloyd, 1992). Because self-incompatibility (SI) is regulated by S-genes, plant scientists have tended to think of SI as a qualitative trait of the breeding systemÐif a species has SI, it is an obligate outcrosser. In reality, however, SI is often a quantitative and phenotypically plastic trait. That is, many species exhibit marked phenotypic variation in the expression of SI that is often in¯uenced by environmental conditions. For example, the expression of SI is known to vary (a) with the action of speci®c S-alleles (e.g. weak and strong alleles); (b) with the expression of modi®er genes; (c) with the composition of the pollen load ( pure self-pollen vs. mixtures of self and cross pollen); (d) with external environmental conditions such as temperature; and (e) with internal stylar conditions such as the age of the ¯ower, time # 2000 Annals of Botany Company

212

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides

of the year, or presence of developing fruit (see Ascher and Peloquin, 1966; de Nettancourt, 1977; Stephenson and Bertin, 1983; Mulcahy, 1984; Lloyd and Schoen, 1992; Levin, 1996; Vogler, Das and Stephenson, 1998; and references therein). These variations in the expression of SI have many names in the literature (incomplete SI, weak SI, leaky SI, partial SI, partial self-compatibility, pseudo selfcompatibility, pollen prepotency etc.) and probably represent a constellation of related phenomena involving subtle pollen±pistil interactions that a€ect the relative performance of self and cross pollen (i.e. speed of germination and/or pollen tube growth rates). For the most part, variations in the expression of SI have been documented by geneticists and plant breeders who view them as nuisances or curiosities that confound attempts to determine genetic speci®cities in cultivated plants (Ascher, 1984). Evolutionary biologists have only recently begun to explore the potential roles of plasticity and plant-to-plant variation in the expression of SI in the origin and maintenance of the breeding system, especially mixed mating systems (e.g. Stephenson and Bertin, 1983; Becarra and Lloyd, 1992; Lloyd, 1992; Latta and Ritland, 1993; Levin, 1996). However, little is known about the genetic and molecular mechanisms that underlie variations in the strength of SI nor the consequences of such variation for the breeding systems of the plants possessing them. For the past decade, our lab has examined variations in the expression of SI in Campanula rapunculoides L. (Campanulaceae). The goals of this paper are: (a) to document variation in the expression of SI; (b) to review our recently published ®ndings concerning the genetic and environmental causes and population level consequences of variation in the expression of SI; (c) to present our recent advances in the molecular biology of the pistil proteins involved in SI in C. rapunculoides; (d) to demonstrate that there is heritable genetic variation for the expression of SI in C. rapunculoides; and (e) to explore the interrelationships among inbreeding depression, variation in the expression of SI, and the evolution of mixed mating systems. N AT U R A L H I S TO RY A N D F LO R A L B I O LO GY O F CA M PA N U L A R A P U N C U LO I D E S Campanula rapunculoides is a naturalized, perennial herb that grows along roadsides and in open woods in the Northeastern United States and Canada (Rosatti, 1986). It overwinters as a rosette and, in midsummer, each rhizomatous cluster produces one to eight ¯owering shoots (racemes) of 10±60 ¯owers that mature acropetally (bottom upwards). The blue, bell-shaped ¯owers are bumblebee pollinated and protandrous. Thus, at anthesis, the ¯owers are staminate (male phase) and the stigmatic lobes are tightly expressed. After 1±3 d, during which the pollen is removed by the bees, the stigmatic lobes re¯ex and pollen deposition can occur ( female phase). Richardson and Stephenson (1989) demonstrated that pollen removal (either manually by the investigators or by natural pollinators) hastens the onset of the female phase. At any given time during the ¯owering season, an in¯orescence will consist of

( from bottom to top) developing fruits, female phase ¯owers, male phase ¯owers, and ¯ower buds. By recording bumblebee movements within an in¯orescence, we found that bumblebees usually begin foraging at or near the lowest open ¯ower and that there is a strong tendency for the bees to forage upwards on an in¯orescence (85% of the movements by bumblebees within an in¯orescence were to a higher ¯ower on the in¯orescence). Consequently, most bees moved from female phase ¯owers to male phase ¯owers within an in¯orescence. The combination of ¯oral protandry (temporal separation of the sexes within ¯owers) and upward foraging by bumblebees on in¯orescences with acropetal development would tend to promote outcrossing. On the other hand, the occasional downward movement by a bumblebee could result in geitonogamy (between ¯ower self-pollination) and geitonogamy also occurs when a bumblebee moves between in¯orescences of the same genotype (rhizomatous cluster). Moreover, when the pollen is not removed from the ¯owers during the male phase, the fully re¯exed stigmas of older female phase ¯owers come into close proximity of the pollen and autonomous selfpollination sometimes occurs ( pers. obs.). F LO R A L A GE D E P E N D E N T VA R I AT I O N I N EXPRESSION OF SI In a preliminary study in a natural population of C. rapunculoides, a series of cross- and self-pollinations on young ( ®rst day of the female phase) and old ( fourth day of the female phase) ¯owers revealed that: (a) self-pollination of young ¯owers results in very low fruit and seed set; (b) self-pollination of old ¯owers results in a signi®cant increase in fruit and seed set; and (c) fruit and seed set decreases slightly from young to old ¯owers following cross-pollinations (Richardson et al., 1990; Stephenson et al., 1992). In addition, we repeated the series of crossand self-pollinations on young and old ¯owers and harvested the styles 6, 12 or 24 h after pollination. These styles were stained with decolourized analine blue and pollen tube growth was examined using ¯uorescent microscopy. We found that pollen tube length increased signi®cantly from 6 to 24 h. More importantly, we found that cross pollen tubes grew signi®cantly more rapidly (and there was less pollen tube bursting) in the styles of young ¯owers but the di€erences in the growth of cross and self pollen tubes decreased signi®cantly in the styles of old ¯owers (Stephenson et al., 1992). In short, these preliminary studies indicated that the ¯owers of most individuals are self-incompatible when they ®rst open but become more self-fertile with ¯oral age. We have now performed controlled pollinations on young and old ¯owers on more than 120 plants that were collected from two natural populations (Centre Co., PA, and Duanesburg, NY, USA). With only a couple of exceptions, the self compatibility index (SCI; self/cross seed set following controlled pollinations) increased from young ¯owers to old ¯owers. For both populations, we have found continuous variation in the SCI among plants ranging from 0 to 0.60 for young ¯owers and 0.02 to 0.95 for old ¯owers (see Table 1). For many of our studies

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides

213

T A B L E 1. The self-incompatibility index (SCI) for a sample of Campanula rapunculoides plants

Day 1 pollinations Day 4 pollinations

1

3

4

5

0.21 0.26

0.48 0.95

0.12 0.17

0.14 0.45

Plant

6

8

13

14

0.05 0.28

0.09 0.22

0.03 0.07

0.55 0.84

The SCI is the number of selfed seeds produced per hand-pollination divided by the number of cross seeds produced per hand-pollination. Day 1 refers to pollinations on ¯owers during their ®rst day in the female phase while day 4 refers to pollinations on ¯owers during their fourth day of the female phase. Data from the plants used by Vogler et al., 1998.

described below, we arbitrarily de®ned three classes of plants: weak SI plants (about 15% of the total plants), in which the young ¯owers were relatively self-compatible (SCI 4 0.40) and became more so as the ¯owers aged; strong SI plants (about 10% of the total plants), in which the old ¯owers remained highly SI (SCI 5 0.15); and typical breakdown plants (35% of total plants), in which the young ¯owers had strong SI (SCI 5 0.15) and the old ¯owers had weak SI (SCI 4 0.40). The remaining plants had intermediate SI phenotypes. PLASTICITY IN THE EXPRESSION OF SI Because developing fruits can draw resources from later ¯owers in an in¯orescence, they have the potential to alter the expression of self-incompatibility in later developing ¯owers on that in¯orescence (Becarra and Lloyd, 1992; Reinhartz and Les, 1994). In this study, we examined the e€ects of ¯oral position within the in¯orescence, ¯oral age, genotype and prior fruit-set on the strength of the selfincompatibility system in C. rapunculoides (Vogler et al., 1998). By treating these factors in a factorial design, we tested for genetic variation and phenotypic plasticity in selfincompatibility (as main e€ects in the analysis of variance) and as interactions among factors. During the summer of 1994, multiple clones of eight genotypes were propagated by rhizome cuttings and brought to ¯ower in a greenhouse. Eight to ten replicate clones of each genotype were matched for overall size and vigour and split into two treatment groups: FRUIT, in which all ¯owers on the lower third of the raceme were handpollinated with cross-pollen, and NO FRUIT in which all ¯owers on the lower third of the raceme were prevented from being pollinated by removing the stigmatic lobes. In the remaining two-thirds of the raceme, most of the ¯owers of each plant continued to receive the appropriate FRUIT/NO FRUIT treatments, but four±12 ¯owers were selected and marked to receive one of four experimental pollinations: (1) self day 1, selfed on the ®rst day of the female phase; (2) out day 1, outcrossed on the ®rst day of the female phase; (3) self day 4; and (4) out day 4. The order of these four treatments was randomized for each plant, and each pollination was performed on ¯owers that were bracketed by other ¯owers receiving the appropriate FRUIT/NO FRUIT treatment for that plant. Even with that spacing, at least one, and often two replicates could be made for each of the four experimental pollinations per plant. We used fresh pollen from male-phase ¯owers for our pollen sources. Pollen loads

were standardized by the method of Richardson et al. (1990) to deliver approx. 500 pollen grainsÐan amount just sucient to produce a full complement of seeds when cross-pollen is used. Capsules were allowed to ripen over a 1 month period; they were then collected and the seeds counted. We found that ¯oral position within an in¯orescence had no e€ect on the strength of SI but we did ®nd that clones in the NO FRUIT treatment had a stronger expression of SI on young ¯owers and a weaker expression of SI on old ¯owers than clones in the FRUIT treatment [i.e. there was a signi®cant treatment (FRUIT/NO FRUIT)  ¯oral age interaction] (Vogler et al., 1998). Because the presence of few or no developing fruit on a plant would be indicative of low pollinator activity in nature, the weaker expression of SI on old ¯owers on the NO FRUIT clones is likely to promote selfed seed set under conditions of pollinator scarcity. Because seed set following cross-pollinations on old ¯owers was greater on the clones in the NO FRUIT treatment than on clones in the FRUIT treatment, we suspect that the increase in self seed on the NO FRUIT plants was due to a greater longevity of the ovules which permitted the slower growing self-pollen to achieve fertilization. Finally, data analyses revealed that plant genotype explained a signi®cant amount of the total variation in the expression of SI and that there was a signi®cant genotype  treatment interaction. Although we used only eight genotypes in the study, these ®ndings strongly indicate that there is broadsense heritability for the magnitude of SI and that SI is a phenotypically plastic trait upon which natural selection could operate. G E NE T I C S A N D M O L E C U L A R B A S I S O F S I I N C . R A P U N C U LO I D E S Our preliminary studies of SI in C. rapunculoides indicated that SI was likely to be gametophytic and that rejection (and/or retardation of growth) occurred in the styles. In order to determine if C. rapunculoides produced S-locus glycoproteins such as those in the Solanaceae, we gathered three types of preliminary evidence. We found that pistil extracts run on SDS±PAGE had an abundant protein (molecular weight approx. 32 kD) in 1-d-old pistils that was still present in 4-d-old pistils, but was not present in extracts from leaf and bud samples (Fig. 1). We found that the 32 kD pistil protein cross-reacted with antibodies to Petunia S-locus glycoproteins ( provided by T-h Kao) and that this protein exhibited RNase activity (see Blank, Sugiyama and

214

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides

FIG. 1. Spatial and developmental regulation of 32 kD putative S-protein in C. rapunculoides. The approx. 32 kD protein is absent in anther (AN) and leaf (LE) tissues and it is not expressed in the pistils from green or purple buds (GB, PB). The protein ®rst appears in pistils in the early male (EM) phase and increases in quantity through late male (LM), early female (EF) phase and is greatest in late female (LF) pistils.

Dekker, 1982 for technique). Although preliminary, these three types of evidence indicate that the molecular properties of the 32 kD protein in C. rapunculoides are similar to those reported for the S-glycoprotein in the Solanaceae (Stephenson et al., 1992). Recently, we found that the 32 kD RNase protein was localized or more abundant (depending upon the genotype) in the top half of the style.

Isoelectric focusing (IEF) gels revealed a highly abundant and polymorphic pistil protein with a similar (and high pH 8.7±9.5) isoelectric point to the Solanaceous Sglycoproteins, and F1 and F2 plants possessed protein bands that co-migrated with the parental bands. More importantly, 2-dimensional gels revealed that the abundant 32 kD protein is the same one that is polymorphic and that it has a high isoelectric point (Fig. 2). Furthermore, we have shown that these proteins have RNase activity (Fig. 3). In short, we have identi®ed a protein that co-segregates with the S-phenotype whose molecular weight, relative abundance, temporal and spatial distribution, RNase activity, and allelic diversity are similar to those reported for S-locus glycoproteins in the Solanaceae. Currently, we are attempting to isolate, amplify, clone and sequence S-alleles from C. rapunculoides. To date S-RNase based SI has been identi®ed in the Solanaceae (the most well studied), Scrophulariaceae, Rosaceae, and now in the Campanulaceae. In order to further explore the heritability of SI and its breakdown, we fully characterized the strength of SI in young and old ¯owers on 35 plants that were originally obtained from one natural population. From these plants,

FIG. 2. Two-dimensional gel of total pistillate protein in C. rapunculoides revealing presence of a highly abundant and polymorphic protein of approx. 32 kD molecular weight. There appear to be four alleles (isoforms) of the protein with isoelectric points between 8.7 and 9.3 ( four arrows on right hand side). Smaller arrow on left identi®es a marker protein at MW 37 kD and pI 5.2.

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides

215

FIG. 3. PAGE-gel stained for RNase activity in 17 individuals of C. rapunculoides. These RNases fall within 30±35 kD. Each individual has between three and seven (usually six) RNase-bands, and the bands are highly polymorphic. Analysis of RNase-activity in IEF gels reveals that one to two of these bands have a low pI (i.e. are acidic), suggesting that individuals have from two to four high pI (i.e. basic) RNase proteins which are the putative S-alleles.

we selected 20 genotypes that represented the range of SI phenotypes (weak SI, strong SI and typical breakdown plants) and randomly crossed them to produce 45 families. From each of these families we grew ®ve progeny and, on each progeny, we made a series of cross- and selfpollinations on young (day 1 in the female phase) and old (day 4 in the female phase) ¯owers. We counted the seeds in the resulting fruits and determined the self compatibility index for young and old ¯owers and the breakdown in SI (SCI of old ¯owers minus the SCI of young ¯owers). To estimate the heritability of the strength of SI on young and old ¯owers and the breakdown in SI, we used the slope from a regression of the mean progeny values (SCI of young ¯owers, SCI of old ¯owers, or SCI of old minus SCI of young ¯owers) on the mid-parent values (Falconer, 1989). Our ®ndings reveal that there is signi®cant narrow sense heritability for the magnitude of SI on young and old ¯owers and for the age dependent breakdown in SI (Figs 4±6). Together with the phenotypic plasticity study, these ®ndings indicate that there is a genetic basis to variation in the expression of SI and that natural selection could act on this variation. Currently, experiments are underway to determine the number of modi®er genes responsible for variations in the expression of SI and whether the modi®er genes are linked (including the possibility of weak and strong S-alleles) or unlinked to the S-locus.

breakdown of self-incompatibility can provide reproductive assurance. If both self and outcross pollen are deposited onto a stigma (as may commonly occur with geitonogamy) the consequences will depend on the ability of the pistil to discriminate against self pollen and the presence of any outcross pollen that shares one or both S-alleles with that plant. In small populations with limited S-allele diversity even random mating patterns may result in some amount of incompatible cross-pollen being deposited on stigmas (Goodell et al., 1997). If the pistil can discriminate against incompatible/self pollen by reducing fertilization success, seeds will be sired disproportionately, or exclusively by outcross pollen. If, however, the breakdown of selfincompatibility allows both self and outcross pollen to fertilize ovules (and there is not complete selective abortion of selfed progeny) then the o€spring will be the product of mixed mating. In this study (Vogler and Stephenson, 2000), equal mixtures of self and cross pollen were deposited onto young and old pistils and the paternity of each progeny was scored using isozyme markers. We found that the progeny from young ¯owers could not be distinguished from expectations of pure outcrossing but that the progeny from old ¯owers included a mixture of selfed and outcrossed seeds. However,

CO N S E Q U E N C E S O F VA R IAT I O N A N D PLASTICITY IN SI ON THE BREEDING SYSTEM The studies reported above have revealed both environmental and genetic variation in the strength of SI. In another series of studies, we have investigated the consequences of this variation in SI on the breeding system of C. rapunculoides. In the ®rst study, we examined the interrelationships among the pollen load, the age dependent breakdown in SI, and the breeding system. We reasoned that the immediate consequences of a breakdown of selfcompatibility will vary with the type of pollen deposited (i.e. self pollen alone or self pollen in competition with outcross pollen) and the ability of the pistil to discriminate against self pollen (Stephenson and Winsor, 1986). If only self pollen is available, as might occur in isolated plants or in populations with low pollinator availability, the

FIG. 4. Heritability of day 1 self-compatibility index (SCI) in C. rapunculoides. The mean value of the progeny day 1 SCI is regressed on the mid-parental day 1 SCI index. The linear regression equation is y ˆ 0.103989 ‡ 0.402380 x. The slope of the regression is an estimate for the narrow sense heritability.

216

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides

FIG. 5. Heritability of day 4 self-compatibility index (SCI) in C. rapunculoides. The mean value of the progeny day 4 SCI is regressed on the mid-parental day 4 SCI index. The linear regression equation is y ˆ 0.172100 ‡ 0.383241 x. The slope of the regression is an estimate for the narrow sense heritability.

FIG. 6. Heritability of the breakdown in self-incompatibility (SI) from day 1 to day 4 in C. rapunculoides. The mean value of the breakdown in progeny degree of SI from day 1 to day 4 is regressed on the midparental value of the breakdown. The linear regression equation is y ˆ 0.129248 ‡ 0.375793 x. The slope of the regression is an estimate for the narrow sense heritability.

the proportion of selfed seeds in the old ¯owers was signi®cantly less than that expected based upon seed set following pure self and pure cross pollinations. It appears that self pollen is discriminated against, even on old pistils. Consequently, the breakdown in SI in C. rapunculoides enhances the potential for sel®ng in the absence of cross pollen but does not substantially reduce the potential for outcrossing when cross and self pollen are both present on a stigma (Vogler and Stephenson, 2000). In the second study, we experimentally varied pollinator availability (access of the pollinators to the plants). Other investigators have found that pollinator availability in natural populations of plants can vary within years, between years and among populations (see Stephenson et al., 1995 for review). Consequently, this study examines an environmentally relevant variable and it di€ers from our previous studies in that it relies on the foraging of natural pollinators and on natural patterns of pollen deposition (mixed and pure pollen loads, variations in the ages of ¯owers at the time of pollen deposition etc). Clones of 20 plants with alternative alleles at the GPI locus [18 with one allele, two with the alternative allele (the focal plants)] were grown in 2 m  2 m  2 m shadecloth cages in the ®eld and exposed to pollinators for 1 h d ÿ1, 24 h every third day, or open pollinated. The three experimental treatments were created by rolling up the sides of the cages for the appropriate amount of time each day. We harvested the mature fruits, counted the seeds, and scored the paternity (self or outcross) of a sample of approx. 100 seeds from each focal plant. The clones for this study were created by repeatedly dividing the rootstalks and then synchronizing the ¯owering schedule by providing an appropriate cold treatment. Our ®ndings from the summers of 1997 and 1998 indicate that the outcrossing rate varied signi®cantly with treatment (open pollination cages 4 1 h cages 4 cages open every third day in both years; 1997 outcrossing rate ˆ 0.77 open cages, 0.70 1 h cages, 0.35 open every third day; 1998 outcrossing rate ˆ 0.55, 0.50, 0.40,

respectively) using plants with the typical breakdown phenotype. In 1997, the weaker SI phenotype set signi®cantly more seeds across the three treatments and in both years there was a signi®cant e€ect of ¯oral position on the outcrossing rate. In the ®rst year, the outcrossing rate increased from the lowest to the highest positions on the in¯orescences and, in year 2, the outcrossing rate was highest in the middle of the in¯orescence (at the peak stage of ¯owering). This study reveals that the breeding system (outcrossing rate) of individual plants can vary with an environmental condition ( pollinator availability) that is known from many studies to vary by year and location. In short, plasticity in the SI system of C. rapunculoides translates into plasticity in the breeding system. In the third study, the consequences of variation in SI on progeny vigour (inbreeding depression) were examined in a multigenerational study (Vogler, Filmore and Stephenson, 1999). Controlled crosses created a range of inbreeding coecients (0, 0.25, 0.50, 0.75) in families derived from strong and weak SI phenotypes. A sample of the progeny from each of these families was scored for survivorship ( percentage germination, survivorship to the rosette stage and survivorship to ¯owering) and reproductive output ( ¯ower number and seeds per fruit). We found that ®tness declined signi®cantly over the range of inbreeding coecients, that the decline in ®tness was less severe for families derived from weak SI phenotypes than for families derived from strong SI phenotypes ( perhaps indicating a prior history of inbreeding and a purging of some of the genetic load), and that inbred maternal plants had a signi®cant non-genetic e€ect on the performance of their progenyÐ even when we compared progeny with the same (0 and 0.5) coecient of inbreeding (Vogler et al., 1999). Our ongoing studies of the progeny from these same families indicates that inbreeding also adversely a€ects the male function of C. rapunculoides ( pollen production, pollen viability, speed of germination and pollen tube growth rates).

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides DISCUSSION Perhaps no area of plant population biology has received more theoretical and empirical attention than the evolution of breeding systems. Fisher (1941) argued that genetic modi®ers that promote self-fertilization without reducing cross-fertilization through the male ( pollen) function should increase due to the transmission of genes via pollen to both selfed and outcrossed progeny. Theoretical studies that assume independence between the level of inbreeding depression and modi®ers of the breeding system (sel®ng rate) predict that the viability of outbred o€spring must exceed the viability of the inbred o€spring two-fold in order to counteract this transmission advantage (e.g. Kimura, 1959; Nagylaki, 1976; Maynard Smith, 1977; Charlesworth, 1980b; and others). Pollen discounting (reductions in outcrossing through the male function due to enhanced sel®ng) and seed discounting (reductions in the production of outcrossed seed due to enhanced sel®ng) promote outcrossing by reducing the minimum levels of inbreeding depression necessary to maintain outcrossing (see Nagylaki, 1976; Holsinger et al., 1984; Lloyd, 1992; Harder and Wilson, 1998). Most of the models that assume independence between the level of inbreeding depression and modi®ers of the breeding system (sel®ng rate) exclude the possibility of mixed mating systems (stable mixes of sel®ng and outcrossing over evolutionary time) (but see Uyenoyama, 1986; Holsinger, 1986 for special exceptions). Models that permit the level of inbreeding depression to change with the breeding system also rarely permit the evolution of mixed mating systems (Maynard Smith, 1977; Lande and Schemske, 1985; Campbell, 1986; Damgaard, Couvet and Loeschcke, 1992). In short, there was a gap between what is (mixed mating systems are not uncommon in nature) and what ought to be (theoretically). To bridge this gap, Holsinger (1988) and Charlesworth and Charlesworth (1990) speculated that the failure of simple models to predict the range of mating patterns observed in nature may re¯ect a violation of the assumption of independence between the genetic modi®ers of the breeding system and the causes of inbreeding depression (viability loci). A series of theoretical studies investigating the nature and magnitude of genetic associations that evolve between loci in¯uencing o€spring viability and modi®ers of SI (e.g. Uyenoyama, 1988) and self-fertilization (Uyenoyama and Waller, 1991a,b,c) conclude that inbreeding depression should not be de®ned at the level of the population but at the level of families (individual plants and their o€spring) and that the associations between the loci determining inbreeding depression and modi®ers of the breeding system (such as SI) together determine evolutionary shifts in the breeding system. Another group of theoreticians have recently attempted to explain the range of mating systems found in nature by including some component of reproductive assurance in their models for the evolution and maintenance of mating systems (e.g. Schoen and Lloyd, 1984; Lloyd, 1992; Holsinger, 1996; Harder and Wilson, 1998). These studies focus attention on the various types of self-pollination (autogamy, pollinator mediated within ¯ower pollen transfer, geitonogamy), the timing of

217

self-pollination (before, during and after opportunities for outcrossing), and the relative performance (speed of germination and pollen tube growth rates) of self and cross pollen in the pistils of ¯owers. Our greenhouse studies of inbreeding depression in C. rapunculoides (Vogler et al., 1999) reveal that the ®tness of a selfed seed ( f ˆ 0.5) is, on average, only 10% of that of an outcrossed seed when one considers only survivorship to ¯owering and reproductive output through the female function. It is reasonable to assume that inbreeding depression would be greater under ®eld conditions than under greenhouse conditions (e.g. Wolfe, 1993; Carr and Dudash, 1995). Moreover, our preliminary evidence indicates that sel®ng also has an adverse e€ect on the male function of Campanula which would further reduce the value of selfed progeny. Even though we (Vogler et al., 1999) found signi®cant variation in the level of inbreeding depression among families, the transmission advantage of selfed seeds is unlikely to outweigh the disadvantages of sel®ng (inbreeding depression, pollen and seed discounting) when seed production is not limited by the availability of outcross pollen. Campanula rapunculoides has at least three mechanisms that avoid sel®ng and promote outcrossing. First, C. rapunculoides is protandrousÐthe temporal separation of the male and female phases of each ¯ower decreases the opportunities for autogamy and pollinator mediated transfer of self pollen within ¯owers. Secondly, the combination of protandry, acropetal in¯orescence development and the foraging behaviour of the bumblebees (the upward movement of bees from female to male phase ¯owers within each in¯orescence) reduces the amount of self pollen that is transferred within in¯orescences. These mechanisms not only avoid the adverse e€ects of sel®ng on progeny vigour but they also avoid pollen and seed discounting. Nevertheless, geitonogamy undoubtedly occurs when the pollinators make the occasional downward movement within an in¯orescence or move between in¯orescences of the same genotype. Thirdly, when self pollen is deposited onto a stigma of a young ¯ower, an S-RNase based SI system prevents or greatly reduces the number of self-fertilizations. When a mixture of self and cross pollen is deposited onto the stigmas of older ¯owers, the cross pollen is signi®cantly more likely to achieve fertilization than the self pollen due to di€erences in the growth rates of self and cross pollen tubes. [It should be noted, however, that while SI systems can avoid inbreeding depression and seed discounting, the pollen deposited onto self stigmas is unavailable for outcrossing (i.e. it is discounted).] Together, these mechanisms ( protandry, protandry/foraging behaviour, SI) promote outcrossing and avoid the adverse consequences of sel®ng when outcross pollen does not limit seed production. However, when seed production is limited by the availability of cross pollen on the stigmas (as would occur when pollinators are scarce, or when the size of the Campanula population is small, or when founding populations consist of only a few S-alleles) plasticity in the SI system of C. rapunculoides allows sel®ng to occur. Our studies reveal that the SCI increases (SI breaks down) with ¯oral age (e.g. Richardson and Stephenson, 1989;

218

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides

Stephenson et al., 1992; Vogler et al., 1998, 1999). Consequently, if a ¯ower is not outcrossed when it is young, it will permit at least some self-fertilization rather than abscising without fruit set. Our studies also reveal that the amount of self seed that can be produced on old ¯owers increases when few fruits are developing on a raceme (when pollination of the earlier ¯owers on a raceme has been inadequate). Finally, when pollinators are scarce (when pollen has not been removed from male phase ¯owers and when cross pollen has not been deposited onto the stigma), the fully re¯exed stigmas of old female phase ¯owers are in close proximity to the left over pollen and may autonomously self pollinate. In short, self-fertility in C. rapunculoides is delayed until after most opportunities for outcrossing have occurred. Consequently, it has, in the words of Becarra and Lloyd (1992), a breeding system that is the best of both worldsÐthe advantages of outcrossing when outcross pollen is available and the higher seed set of a selfer when outcross pollen is unavailable. In addition to environmental e€ects on the expression of SI ( ¯oral age, prior fruit production) we also found signi®cant additive genetic variation among plants for the strength of SI in young and old ¯owers and for the amount of breakdown in SI with ¯oral age (Figs 4±6). These ®ndings indicate that natural selection could operate on the variation present in the expression of SI within the natural populations that we have studied. Currently, we are conducting experiments in which we vary pollinator access (availability) using arti®cial populations in shadecloth cages to determine the ®tness (reproductive output through the male and female functions) of weak and strong SI-phenotypes under di€erent environmental conditions. Finally variation in the expression of SI is common (see de Nettancourt, 1977; Stephenson and Bertin, 1983; Mulcahy, 1984; Lloyd and Schoen, 1992; Levin, 1996; Vogler et al., 1998 and references therein). In fact, Ascher (1984) claims that variation in the expression of SI occurs in virtually all species in which SI has been seriously examined. Although it may be useful for plant breeders and molecular geneticists to ignore or avoid this variation in their studies, it is becoming increasingly clear that SI is not a qualitative trait de®ning a species breeding system. Rather, SI should be viewed as a more subtle component of the breeding system. For example, our study that used shadecloth cages to permit open access by the pollinators, access for 1 h per day, or access on every third day revealed that the outcrossing rate on the same genotypes replicated across the three environments varied from nearly 0.35 to 0.77. Moreover, variations in the expression of SI due to genotype (such as heritable variation in genetic modi®ers of SI), environment (such as ¯oral age, prior fruit production, temperature, etc), and genotype  environment interactions are likely to be key, and often overlooked, components in the evolution of many mixed mating systems. In short, our inability to develop models for the evolution of mating systems that are capable of generating the range of mating systems found in nature may be due to the fact that plasticity in the mating system is selected in nature rather than a single (sel®ng, outcrossing, or mixed) mating system.

AC K N OW L E D G E M E N T S We thank Christine Difolco, Kevin Filmore, Chandreyee Das, Cortney Springstead, Lisa Copper, Tom Nagel, and Christina Klescz for ®eld, greenhouse and lab assistance, Tony Omeis and his sta€ at the Buckhout Greenhouse, and Bob Oberheim and his sta€ at The Pennsylvania State University Agricultural Experiment Station at Rock Springs, PA, USA. This research was supported by NSF grant DEB 95-27739 to A.G.S.

L I T E R AT U R E C I T E D Ascher PD. 1984. Self-incompatibility in Petunia. Monographs on Theoretical and Applied Genetics 9: 92±110. Ascher PD, Peloquin SJ. 1966. E€ect of ¯oral aging on the growth of compatible and incompatible pollen tubes in Lilium longi¯orum. American Journal of Botany 53: 99±102. Baker HG. 1955. Self-compatibility and establishment after `longdistance' dispersal. Evolution 47: 125±135. Becarra JX, Lloyd DG. 1992. Competition dependent abscission of self-pollinated ¯owers of Phormium texax (Agavaceae): a second action of self-incompatibility at the whole ¯ower level?. Evolution 50: 892±899. Blank A, Sugiyama RH, Dekker CA. 1982. Activity staining of nucleolytic enzymes after sodium dodecyl sulfate±polyacrylamide gel electrophoresis: use of aqueous isopropanol to remove detergent from gels. Analytical Biochemistry 120: 267±275. Campbell RB. 1986. The interdependence of mating structure and inbreeding depression. Journal of Theoretical Biology 30: 232±244. Carr DE, Dudash MR. 1995. Inbreeding depression under a competitive regime in Mimulus guttatus: consequences for potential male and female function. Heredity 75: 437±445. Charlesworth B. 1980a. Evolution in age-structured populations. Cambridge: Cambridge University Press. Charlesworth B. 1980b. The cost of sex in relation to the mating system. Journal of Theoretical Biology 24: 655±671. Charlesworth D, Charlesworth B. 1990. Inbreeding depression with heterozygote advantage and its e€ect on selection for modi®ers changing the outcrossing rate. Evolution 44: 870±888. Damgaard CD, Couvet D, Loeschcke V. 1992. Partial sel®ng as an optimal mating strategy. Heredity 69: 289±295. Darwin C. 1876. The e€ects of self and cross fertilisation in the vegetable kingdom. New York: Appleton and Company. de Nettancourt D. 1977. Incompatibility in angiosperms. Berlin: Springer-Verlag. Falconer DS. 1989. Introduction to quantitative genetics, 3rd edn. New York: Longman. Fisher RA. 1941. Average excess and average a€ect of a gene substitution. Annals of Eugenics 11: 53±63. Goodell K, Elam DR, Nason JD, Ellstrand NC. 1997. Gene ¯ow among small populations of a self-incompatible plant: an interaction between demography and genetics. American Journal of Botany 84: 1362±1371. Harder LD, Wilson WG. 1998. A clari®cation of pollen discounting and its joint e€ects with inbreeding depression on mating system evolution. The American Naturalist 152: 684±695. Holsinger KE. 1986. Dispersal and plant mating systems: the evolution of self-fertilization in subsidivided populations. Evolution 40: 405±413. Holsinger KE. 1988. Inbreeding depression doesn't matter: the genetic basis of mating system evolution. Evolution 42: 1235±1244. Holsinger KE. 1996. Pollination biology and the evolution of mating systems in ¯owering plants. Evolutionary Biology 29: 107±149. Holsinger KE, Feldman MW, Christiansen FB. 1984. The evolution of self-fertilization in plants: a population genetic model. The American Naturalist 124: 446±453. Kimura M. 1959. Con¯ict between self-fertilization and outbreeding in plants. Annual Report of the National Institute of Genetics, Japan 9: 87±88.

Stephenson et al.ÐSelf-incompatibility in Campanula rapunculoides Lande R, Schemske DW. 1985. The evolution of self-fertilization and inbreeding depression in plants. I. Genetic models. Evolution 39: 24±40. Latta RG, Ritland K. 1994. The relationship between inbreeding depression and prior inbreeding among populations of four Mimulus species. Evolution 48: 806±817. Levin DA. 1996. The evolutionary signi®cance of pseudo-self-fertility. The American Naturalist 148: 321±332. Lloyd DG. 1992. Self- and cross-fertilization in plants. II. The selection of self-fertilization. International Journal of Plant Science 153: 358±369. Lloyd DG, Schoen DJ. 1992. Self- and cross-fertilization in plants. I. Functional dimensions. International Journal of Plant Science 153: 358±369. Maynard Smith J. 1977. The sex habit in plants and animals. Berlin: Springer Verlag, 265±273. Mulcahy DL. 1984. The relationships between self-incompatibility, pseudo-compatibility and self-compatibility. Toronto: Academic Press, 229±235. Nagylaki T. 1976. A model for the evolution of self-fertilization and vegetative reproduction. Journal of Theoretical Biology 58: 55±58. Reinartz JA, Les DH. 1994. Bottleneck-induced dissolution of selfincompatibility and breeding system consequences in Aster furcatus (Asteraceae). American Journal of Biology 81: 446±455. Richardson TE, Stephenson AG. 1989. Pollen removal and pollen deposition e€ect the duration of the staminate and pistillate phases in Campanula rapunculoides. American Journal of Botany 76: 532±538. Richardson TE, Hrincevich A, Kao T-h, Stephenson AG. 1990. Preliminary studies into age-dependent breakdown of self-incompatibility in Campanula rapunculoides: seed set, pollen tube growth and molecular data. Plant Cell Incompatibility Newsletter 22: 41±47. Rosatti TJ. 1986. The genera of Sphenocleaceae and Campanulaceae in the Southeastern United States. Journal of the Arnold Arboretum 67: 1±64. Schemske DW. 1978. Evolution of reproductive characteristics in Impatiens (Balsaminaceae): the signi®cance of cleistogamy and chasmogamy. Ecology 59: 596±613. Schoen DJ, Lloyd DG. 1984. The selection of cleistogamy and heteromorphic diaspores. Biological Journal of the Linnean Society 23: 303±323. Schoen DJ, Morgan MT, Bataillon T. 1996. How does self-pollination evolve? Inferences from ¯oral ecology and molecular genetic

219

variation. Philosophical Transactions of the Royal Society, London 351: 1281±1290. Solbrig OT. 1976. On the relative advantages of cross- and selffertilization. Annals of the Missouri Botanical Garden 63: 262±276. Stebbins GL. 1957. Self-fertilization and population variability in the higher plants. The American Naturalist 91: 337±354. Stephenson AG, Bertin RI. 1983. Male competition, female choice, and sexual selection in plants. New York: Academic Press, 109±149. Stephenson AG, Winsor JA. 1986. Lotus corniculatus regulates o€spring quality through non-random fruit abortion. Evolution 40: 453±458. Stephenson AG, Quesada MR, Schlichting CD, Winsor JA. 1995. Consequences of variation in pollen load size. Monographs in Systematic Botany 53: 233±244. Stephenson AG, Winsor JA, Richardson TE, Singh A, Kao T-h. 1992. E€ects of style age on the performance of self and cross pollen in Campanula rapunculoides. New York: Springer-Verlag, 117±121. Uyenoyama MK. 1986. Inbreeding and the cost of meiosis: the evolution of sel®ng in populations practicing biparental inbreeding. Evolution 40: 388±404. Uyenoyama MK. 1988. On the evolution of genetic incompatibility systems. II. The American Naturalist 131: 700±722. Uyenoyama MK, Waller D. 1991abc. Coevolution of self-fertilization and inbreeding depression. I, II, III. Theoretical and Population Biology 40: 14±46; 47±77; 173±210. Vogler DW, Stephenson AG. 2000. The potential for mixed mating in a self-incompatible plant. International Journal of Plant Science (in press). Vogler DW, Das C, Stephenson DG. 1998. Phenotypic plasticity in the expression of self incompatibility in Campanula rapunculoides. Heredity 81: 546±555. Vogler DW, Filmore K, Stephenson AG. 1999. A comparison of inbreeding depression in plants derived from strong and weak self incompatibility phenotypes in Campanula rapunculoides L. Journal of Evolutionary Biology 12: 483±491. Waller DM. 1979. The relative cost of self- and cross-fertilized seed in Impatiens capensis (Balsaminaceae). American Journal of Botany 66: 313±320. Wolfe LM. 1993. Inbreeding depression in Hydrophyllum appendiculatum: role of maternal e€ects, crowding and parental mating history. Evolution 47: 374±386.