Genetic Control of Self-Incompatibility in Centromadia (Hemizonia) Pungens Subsp. Laevis (Madiinae, Asteraceae)

The presence of se lf-incompatibility was tested in Centroma dia pun gens subsp. laevis and (he gene tic basis of the se lf-incompatib ility respon se was ex plored using crossing stud ies. We performed full d iallel c ross ing experiments amo ng 10 ind ividuals from one natural population and four F, families . We observed a strong se lfinco mpaubility resp onse in a ll indi viduals tested . with a significant d ifference in seed se t between se lfed and outcrosse d maii ngs . Most pairw ise rnat ings amo ng parental plants were co mpatib le. with so me nonreciprocall y incom pat ible matin gs (i .e.. the rnatin gs we re suc cessfu l in one direction . but not the other), and only one recipro cally incompatible mating. The full diallel cross ing s tudies among sibs in the four F, families showed two major compatibility c lasses. Th ese res ults are consistent with a sing leloc us spo rophytic se lf-incompatibility system in this spec ies .


INTRODUCTION
The majority of flow ering plants are hermaphroditi c, having both male and female floral organs. The clo se proximity of the sexual organs allows for self-pollination by pollen tran sfer within the flower. While selfpollination is common among plants, man y mechanisms that promote outcrossing have evol ved , impl ying that there is a n e volutionary advantage in av oiding self-fertilization (Richards 1997; Steinbachs and Holsi nge r 1999 ). Among the many mechanisms that promote out crossing, including dioecy, monoecy, dist yly, and oth er morphol ogical adaptations, self-incompatibility (SI) is of interest for several reasons. First, SI is a genetically regulated self-recognition system. Second, the genetic and molecular bases of SI are understood in some, but not all, systems. Third, SI has a wide phylogenetic distribution in the Angiosperms, with several different mechanisms, indicating multiple evolutionary origins (C harles worth and Charlesworth 1979;de Nett ancourt 200 I;Lundq vist 1990a. b ;Richard s 1997 ).
Although the phenomenon of SI has been obser ved for over 200 yea rs (Kolreuter 1763 ) the genetics controlling it were not discovered until the 1920's (Ea st and Man gel sdorf 1925 ; Prell 1921) and not until the last 50 years have there been exhaustive pollination studies to e luc ida te the mechanisms of the different forms of SI. Self-incompatibility systems can be categorized into two basic types, gametophytic SI (GSI) and sporophytic SI (SSI). Gametophytic SI is widespread in the Angiosperms, occurring in at least 15 families (Richards 1997 ;Newbigin et al. 1994). Crossing success under gametophytic control is determined by the haploid genotype of the individual pollen grain and the diploid genotype of the pist illate parent . Crossing a bility in sporophytic systems, on the oth er hand, is controlled by the diploid genotype of each parent. Thus, a success fu l cro ss can only occur when the phenot ype , which may be different than the dipl o id genot ype due to dominance interactions between allel es, of each parent is different. Sporophyti c SI is known in six plant families , including Asterace ae , Betulaceae, Caryophyllaceae, Convol vula ceae, Ster cul iace ae, but has been e xtensively studied at the molecular level in only one family, Brassicaceae (Kao and McCubbin 1996 ).
The genetic control of the SI respon se is relatively eas y to characterize by crossing studies of full-s ib families. In any F , family segregating for four alleles , single-l ocus GSI and SSI can be distin gui shed by the cross ing behavior of the siblings. One key characteristic of GSI sys tems is the presence of semi-compatible crosses. In GSI , if the parents share one allele, 50 % o f pollen grains will be capable o f effecting pollination . Gametophytic systems are also c ha rac terized by the pre sence of four equally common cross ing type s among the F , individuals and the ab sence of nonreciprocal crosses (Goodwillie 1997; Rich ard s 1997 ). In SSI sys te ms. there are no semi-compatibl e cross es. but there may be two, three or four crossing type s among F, individuals, due to dominance relationships amo ng alle les . Also. nonreciprocal crosses are common in SSI sibships, because of dominance relationships amo ng alleles (Richa rds 1997). Se ver al cross ing studies within Asteraceae support the pre sen ce of sing le-locus SSI. the first studies being of Cosmos bipinnatus (Crowe 1954 ), Crepis foetida (Hughes and Babcock 1950), and Parthenium argen-tatum (Gerstel 1950). Additional studies within the family include: Hym enoxys acaulis var. glabra (DeMauro 1993), Aster furcatus (Les et al. 1991), Senecio squalidus (Hiscock 2000a), Calotis cuneifo lia (Davidson and Stace 1986), Helianthus spp. (Desrochers and Rieseberg 1998), and Rutidosis leptorrhynchoides (Young et al. 2000). However, several recent studies have suggested a breakdown of the SSI system, or the presence of additional loci controlling the response, suggesting that the SSI system within Asteraceae may not be as simple as previously thought (Les et al. 1991;DeMauro 1993;Hiscock 2000b). These results serve as reminders that exhaustive studies are necessary in order to get a complete picture of the SI mechanism within any given family. Here we pre sent pollination data for Centromadia pungens (Hooker & Arnott) Greene subsp. laevis (D. D. Keck) B. G . Baldwin, which will be used as the model system for ongoing molecular and genetic studies of SI in Asteraceae subtribe Madiinae.

Plant Material
Centromadia pungens subsp. laevis, formerly Hemironia pungens (Hooker & Arnott) Greene subsp. laevis D. D. Keck, is a small annual plant of the South Coast and Peninsular Ranges of Southern California, typically growing in grassland habitats. The inflorescences measure 4 -6 mm and have female ray florets and bisexual (functionally staminate) disk florets.
Parental plants were derived from seed collected from the San Jacinto Valley in Riverside County, California in early July 1996 (Michael Wall 146, RSA). These seeds were placed into long-term storage at the Rancho Santa Ana Botanic Garden for future use in reintroduction and restoration projects (acces sion # 19178). Ten seeds from the RSABG seed storage program were randomly chosen and germinated on agar plates, then transplanted into a steri Ie potting mixture. Plants were grown in a growth chamber at 14 hour days, with daytime temperatures of 26°C and nighttime temperatures of 12°C. Parental plants were numbered 1-10.
Subsequent generation plants were derived from controlled crosses among the parental plants. The achenes were harvested when fully ripe and stored until use . Achenes were germinated and grown as above. F 1 families were numbered 1-4. Individuals in each F 1 family were identified by letters.

Crossand Self-Pollinations
Parental generation. -A full diallel crossing study wa s performed on the ten individuals from the wild population to estimate the number of alleles present in the natural population and generate F] families for further study. To ensure control of pollen movement, disk florets were removed from heads to be used as female parents prior to anthesis . All pollinations were performed after all ray florets had opened and the stigmatic lobes had reflexed. At that point. heads of the male parent that were presenting pollen were brushed onto the stigmatic lobes . This procedure was followed for both cross-and se lf-po llinatio ns. Each cross-pollination was performed at least three times . Each selfpollination was performed at least five times. In addition, two head s per individual were simply bagged to determine the amount of seed set without pollen movement. Heads were removed from the plants after 3-4 weeks and the number of filled ray achenes counted . The percentage of seed set was calculated using the number of ray florets and the number of filled achenes. Filled achenes could be easily distinguished based on their plump appearance and indurate achene wall. Unfilled achenes were thin and papery. Several achenes of each type were dissected to verify the correlation between achene appearance and embryo development.
F J generation.-Full diallel crossing studies were performed on four full-sib families from the parental generation. The four families were chosen to encompass the range of crossing results from the parental generation. Two famili es represent offspring from a reciprocally successful mating in each direction (# I: l oX 3 <;;>; #3: 33 X I <;;», one family represents an additional reciprocally successful mating (#2: 23 X 7 <;;», and one family represents the offspring from a non-reciprocal mating (#4: 7 3 X I <;;>; Fig. 2).

Analysis of Crossing Data
The mean number of filled achenes for self-and outcross pollinations was compared using a o ne-tailed paired t-test as implemented in StatView 5.0.1 (SAS Institute, Cary, NC). The correlation between self and outcrossed seed set and male and female success for individual plants was also calculated using StatView 5.0.1.
The crossing results in the parental generation were also examined to determine the number of S-alleles present in the sample. S-genotypes were assigned to each individual based on crossing phenotype. It wa s assumed that individuals showing nonreciprocal incompatibility shared one S-allele, that dominance hierarchies between alleles were possible, and that all individuals were heterozygous at the S-Iocus. Alleles were defined as recessive if, when shared, the cross wa s nonreciprocally incompatible. For instance, the I X 7 cross is nonreciprocally compatible (i.e., 13 X 7 <;;> is successful, I <;;> X 70 is not). It wa s assumed, therefore, that these two individuals shared an S-allele in common. Moreover, as the cross was successful when individual 1 was the male parent, it was assumed that one allele in individual 1 was recessive in the pollen, and that this allele was shared with individual 7. The total number of alleles in the sample was estimated from the S-genotype assignments.

Parental Generation
The mean number of ray florets scored per head was 13.8 (4.83 standard deviation), with a mean of 56.6% (40.6%) seed set overall. The mean percentage seed set for outcrossed heads was 65.9% (36.1 %). The mean percentage seed set for selfed heads was 0.2% (1.1 %). Outcrossing resulted in significantly greater seed set than selfing for each plant (paired z-test, t = -13.126, P < 0.000 I). There was no significant correlation between outcross seed set and selfed seed set for individual plants (r = -0.121, P = 0.75). There was considerable variation in mean seed set (65.1 % to 88.04%) and mean male success, defined as mean seed set for an individual over all matings made as a pollen parent (52.7% to 91.2%). However, there was no correlation between success as a male parent and as a female parent for individual plants (r = 0.203, P = 0.59). Heads that had been bagged without any pollen movement had no observed seed set.
A frequency histogram for the mean percent seed set for both self and outcross matings is shown in Fig.  I. Each bar represents the percentage of matings of each cross type (i.e., self vs. outcross) with the respective percentage of filled achenes. This frequency distribution was used to assign compatibility types for all matings. All matings were categorized as being compatible, incompatible or indeterminate based on the percentage of filled achenes. Incompatible matings were defined as those that had a mean of less than 10% filled achenes, compatible matings were those that had a mean of greater than 40% filled achenes, and indeterminate matings had between 10% and 40% filled achenes. These limits were defined to include the maximum number of outcrosses as compatible and allow for variation in the success of seed set per individual. Only five crosses fell into the 10-40% range; it is unclear whether these represent "leaky" incompatible crosses or compatible crosses with poor seed set. The distribution of matings was highly skewed, with most outcross matings having high percentages of seed set, and all selfed matings and a minority of outcross matings having very low seed set.
Centromadia pungens subsp. laevis appears to be strongly self-incompatible. There is a significant difference in the mean seed set between self and outcross matings, with selfed matings resulting in virtually no seed set. Outcross matings resulted in variable amounts of seed set. The distribution of seed set for outcross matings is largely bimodal, with a large peak near 100% and a smaller peak near 0% seed set, with few matings showing intermediate values of seed set. These results suggest that the majority of outcross matings are compatible, with a minority that are strongly incompatible. Figure 2 shows the distribution of compatible and incompatible matings among the ten individuals from the natural population. Of the 90 possible outcross

F) Generation
Complete dial leI crossing experiments were performed with four F) families derived from crosses in Tabl e I . All eli c assignments of individuals in the parental generation based on inference from crossing phenotype . The four familie s were chosen to represent the range of crossing results in the parental generation, including both reciprocally compatible and nonreciprocal matings. The sample sizes for the four families vary due to differences in germination rate among them. As the parental generation produced few matings with intermediate percentages of seed set, we were able to score matings in the F, generation simply as compatible or incompatible . Therefore, statistics on the percentage of filled seeds were not recorded for the F I generation . The results of the diallel crossing experiments are presented in Fig . 3-6.
All families show two major compatibility groups, with occasional aberrant results. Most of the aberrant results appear to be related to low seed set in a few individuals. For instance, in family #2, individual s 2D and 2G never set seed regardless of the male parent used (Fig. 4). Family #4, the only F, family resulting from a non-reciprocal cross ( Fig. 2 ; 7 <3 X I «), showed the most aberrant results. In the inferred allelic compositions of each individual , it was assumed that parental individuals I and 7 share an allele in common , either 5) or 52> which is recessive in the male parent. The possibility of homozygous individuals resulting from thi s c ross may explain the aberrant results. Alternatively, the aberrant results may be due to poor Fig . 3. Diallel crossing results for F , famil y # I (1 0 X 3 <:( ) . An expl anation of symbols and shading is given in Fig. 2.  matings, 83 (92.2%) were compatible. Of the incompatible rnatings, most were nonreciprocally incompatible (i.e ., compatible in one direction but not the reverse) . Of the 45 pairs of plants, 40 were reciprocally compatible, four were nonreciprocally incompatible, and one was reciprocally incompatible (Fig. 2) . The assignment of 16 5-alleles to the sample of 10 individuals explains all of the observed mating phenotypes (Table I) . [n one incompatible cross (1 <3 X 9 «), it is clear that individual 9 shares an allele with individual 2, but which allele cannot be determined, so we arbitrarily assigned allele 53 to individual 9 for purposes of discussion. In addition, the se assignments assume that 51 is dom inant to 53' but recessive to 5 13 , and 52 is recessive to 5 16 in pollen in order to explain the nonreciprocal matings.
These allele assignme nts, while relatively arbitrary, indicate that a sample of ten individuals from the natural population contains approx imately 16 5-alleles, indicating significant genetic diversity at this locus. These results are typical for self-inco mpatible taxa , which typically show large numbers of alleles maintained in relatively small populations (Lawrence 2000). By contrast, Brennan et al, (2002) found that a natural population of Senecio squalidus had only six alleles in a natural population. This finding was explained by the presence of a severe population bottleneck at the introduction of this species to Britain. The presence of nonreciprocal crossing results in both the parental and F 1 generations and two com patibility classes in the F 1 sibships are consistent with sing le locus sporophytic control of self-incompatibility. These results are also con sistent with previous findings throughout the phylogenetic span of Asteraceae, which have shown single-locus sporophytic cont rol of SI (Gerstel 1950;Hughes and Babcock 1950;Crowe 1954;Hiscock 2000a). As 51 systems appear to be a family-level trait (with at lea st one exception in the Polemoniaceae; Levin 1993 ; Goodwillie 1997 ; La-Doux and Friar in prep.), all self-incompatible taxa in the Asteraceae may share a similar system (de Nettancourt 200 I). However, there wa s no evidence in these cros sing results of the anomalous compatibility found by Hiscock (2000b) , who found fully or partially compatible crosses in otherwise completely incompatible groups. These results led him to postulate the presence of an additional gamerophytic locus controlling the in- l Eo Fi g. 5. Diallel cro ssing results for F, family #3 (30 x 1 2 ), An e xplanatio n of symbol s a nd shading is given in Fig. 2. compatibility respon se . This pattern has been found in several other self-incompatible species, including erepis foe tida and Hyp ochoeris radicata in the Asteraceae, se veral s pec ies in the Brassicaceae, and at lea st two species in the Caryophyllaceae (Lewis et al. 1988;Lundqvist I990a, 1994;Hiscock 2000b ).
Recent results (Hiscock et al. 2003) suggest that the self-incompatibility systems of the Asteraceae and Brassic aceae may not be homologous at the molecular level , de spite the similarities in phy siology and inheritance. The members of the Asteraceae that have been investigated to date seem to share a similar singlelocus sporo phytic SI syndrome, though with some implications of other modifying loci in additio n to the locus of large effect. Once the gen es res pons ible for the SI res ponse in the Asteraceae are cloned, the molecular, developmental , and/or physiological homologies between the two families, and among taxa within the Asteraceae, can be addressed .