Euphytica 62: 155-169, 1992. 1992 Kluwer Academic Publishers. Printed in the Netherlands.
Inbreeding depression in garden and glasshouse chrysanthemums: germination and survivorship Neil O. Anderson, Peter D. Ascher & Richard E. Widmer
Department of Horticultural Science, University of Minnesota, 1970 Folwell Avenue, St. Paul MN 55108, USA Received 17 April 1991; accepted 10 June 1992
Key words: Dendranthema grandiflora, chrysanthemum, hybrid seed development, inbreeding depression, pseudo-self compatibility, rapid generation cycling, recombinant inbreds, self incompatibility
Summary Sixty-six chrysanthemum (Dendranthema grandiflora) inbred selections, noninbred cultivars and hybrids, and D. makinoi were the base populations from which up to three generations of inbreds were obtained using multiple plant descent. Each parent possessed pseudo-self compatibility (PSC), which allowed seed set following self pollination. Rapid-generation cycling (laboratory seed ripening to heart stage and subsequent embryo rescue) reduced generation time and minimized confounding maternal with zygotic inbreeding depression during post heart-stage seed development. Selection criteria were male/female fertility and PSC. Two stages of the life cycle were chosen to evaluate inbreeding depression: germination (seed development to germination) and survivorship (fertile individuals at anthesis). PSC was environmentally interactive and genotype dependent, causing variable levels of self seed set between and within populations and generations. By the end of the second generation, families from all noninbred cultivars, D. makinoi, and one F1 hybrid were eliminated from the experiment due to self incompatibility and/or inbreeding depression. Postgermination inbreeding depression was severe in several advanced inbreds. Inbred progeny of most F1 or F2 hybrids expressed less or equal amounts of inbreeding depression compared to advanced inbreds. Linear regression coefficients for either germination or survivorship on percent homozygosity were negative. Correlation coefficients between percent germination and survivorship (as a percent of seed set) were highly significant for Minnesota inbreds (r = 0.67, P ~ 0.002) and hybrids (r = 0.67, P -< 0.006). The correlation coefficient was higher when percent germination and survivorship (as a percent of germinated seedlings) were compared (r = 0.95, P-< 0.001).
Abbreviations: F - Fisher's Coefficient of Inbreeding, IBD - Inbreeding Depression, PSC - Pseudo-self Compatibility, RGC - Rapid Generation Cycle, SI - Self Incompatibility
Introduction Inbreeding can be defined as the mating of individuals more closely related to each other than are
random members of an infinitely large population (Allard et al., 1968). Essential consequences of inbreeding are a reduction in genetic variance within families and an increase in genetic variance be-
Scientific Journal Series Paper Number 17,282 of the Minnesota Agricultural Experiment Station
156 tween families (Johannsen, 1926). The effect of inbreeding in normally cross-fertilized organisms is to increase homozygosity (Fowler, 1965a), which results in inbreeding depression (IBD) at one or several critical stages of the life cycle (Darwin, 1876; Schemske, 1983). IBD is defined as a decrease in fitness and vigor resulting from selfing individuals that are normally outcross pollinated, due to the expression of recessive deleterious mutations in homozygotes (Campbell, 1988). Outcross pollination is defined as the mating of individuals at random from an infinitely large population (Allard et al., 1968). The major force maintaining outcrossing thus appears to be substantial IBD in selfed progeny (Fowler, 1965b; Schemske & Lande, 1985). Outcross pollination is the predominant breeding system for species in the polyploid Chrysanthemum complex (Tanaka, 1952). This is enforced by self incompatibility (SI), a genetically based physiological mechanism which acts as a prefertilization barrier when there is a match in the S specificities between pollen and pistil (Ascher, 1976). SI is distributed across all ploidy levels in the chrysanthemum complex (Thorpe, 1940; Tanaka, 1952; Brewer, 1968). Diploid Dendranthema boreale Makino (= D. lavandulaefolium) and Tanacetum cinerariaefolium Schultz Bip. (Trev.), hexaploid D. japonense Nakai, D. grandiflora Tzvelv., octaploid D. ornatum Hemsl., and decaploid D. shiwogiku Shimotomai et Kitamura have been reported to be SI. Genetic analysis of cultivated chrysanthemums (D. grandiflora) revealed the presence of a sporophytic SI system (Mulford, 1937; Crook, 1942; Drewlow et al., 1973) with at least three epistatic S loci (Zagorski et al., 1983). D. grandiflora Tzvelv. (Chrysanthemum morifolium Ramat.) (Anderson, 1987) is usually highly SI, producing no or few seeds after self pollination. However, pseudo-self compatibility (PSC), defined as limited to full seed set following selling in a plant possessing an active SI system, has surfaced at a very low frequency in most chrysanthemum breeding programs (Wu, 1963; Kawase & Tsukamoto, 1966, 1977; Stephens et al., 1983). PSC individuals produce progenies that are either SI or PSC
following self or outcross pollination with SI individuals (Ronald & Ascher, 1975a, 1975b). In most polyploid chrysanthemum species (including D. grandiflora), preferential pairing results in a stable, essentially diploid meiosis, which facilitates sexual reproduction (Dowrick & EI-Bayoumi, 1966; Watanabe, 1977a, 1977b). Similar systems controlling meiotic pairing have been found in other polyploid crops, such as Arena (Rajhathy, 1971) and Lolium (Evans & Macefield, 1972). Since cultivated hexaploid chrysanthemums exhibit disomic inheritance, the disomic rather than the hexasomic equation for F, the coefficient of inbreeding (Kempthorne, 1957), applies to this species. Marshall (1973) proposed that inbreeding may be relatively ineffective for cultivated, allohexaploid chrysanthemums because of fixed heterozygosity; that is, heterozygosity in duplicated loci originating through polyploidy. Fixed heterozygosity is a frequently observed consequence of allopolyploid speciation (Bingham, 1979; Haufler & Soltis, 1986; Soltis, 1986). References to the possibility of IBD in D. grandiflora appear in experiments designed to analyze the inheritance of other traits. In the first report to document SI in chrysanthemums, wild and cultivated species were selfed to produce one to three inbred generations in five years (Niwa, 1931). Selling resulted in decreased height, seed set, germination, and flowering, when compared to outcross pollinations, leading Niwa to conclude that these normally outcrossing species exhibited IBD. Subsequent studies (Mulford, 1937; Tsukamoto et al., 1964; Kawase & Tsukamoto, 1966) also compared the performance of F1 and 11 (first inbred generation) progeny for fertility or flower characters. While there was a general reduction in average I~ performance, the standard errors were large enough to encompass the range observed for the
El. For the past 10-15 years, the University of Minnesota chrysanthemum breeding program has been inbreeding and progeny testing potential parents for producing F 1 hybrid seed cultivars. It is not known, however, the extent to which IBD may impede obtaining fertile, homozygous inbred par-
157 ents. Previous results have demonstrated that inbreeding can give rise to many sublethals that do not lead to direct extinction, i.e. rosetting, partial chlorophyll deficiencies, foliar or floral abnormalities (deformed leaves, apetalous ray florets, etc.), and decreased height or vigor. These would contrast with lethal manifestations of IBD: sterility, nongermination, albinism, and floral initiation and/or development abnormalities. The purpose of this experiment was to investigate whether cultivated chrysanthemum populations (at various levels of inbreeding) expressed IBD (in the strictest sense, i.e. a lethal) during three generations of selfing and whether the IBD response followed the theoretical expectations.
Materials and methods
Populations of inbred selections, noninbred cultivars, and hybrids of D. grandiflora (2n = 6x = 54) from the University of Minnesota garden chrysanthemum breeding program, and D. makinoi (= D. japonicum) (2n = 2x = 18) - a diploid progenitor species - were grown to anthesis under glasshouse conditions (45°N latitude at St. Paul, Minnesota). Noninbred cultivars included both garden ('Daisy White', 'Princess', and 'Yogo') and glasshouse ('Echo') clones. Five plants were selected from each of three populations of D. makinoi (Seed Lot Nos. 2026A, 5832A and 1701M, T. Sakata Seed Co., Yokohama, Japan). Fchybrid, Minnesota Selection 86-N711-1 resulted from a cross between 'Royal Pomp' x bulk pollen from unrelated Minnesota inbred-derived F1 hybrids (Anderson et al., 1990), while 'Autumn Glory' (Seed Lot No. N7387) and Korean-hybrid 'Glorious' (Seed Lot No. 7332A) were commercially available F1 hybrids (with mixed flower colors) introduced by the T. Sakata Seed Company in 1972 and 1962, respectively. 'Petit Point' is an 'F2' hybrid population (Bodgers Seeds, USA) obtained by open pollinating (sibbing) and/or selfing 50,000 'Autumn Glory' F1 hybrids (Lyndon Drewlow, personal communication, 1986). Uniformity for several phenotypic traits suggest-
ed that the commercial hybrids were derived from inbred parents, but the pedigrees of F1-hybrid 'Autumn Glory' and 'Glorious' were proprietary information and the coefficients of inbreeding could not be calculated. Thus, it was assumed that no common ancestor existed in the hybrid pedigree and all FI hybrid values were set at zero. Because it is unknown whether the F2-hybrid 'Petit Point' was derived via selfing (F = 0.5) or full-sib mating (F = 0.25), the value F = 0.25 was chosen for this experiment. Sixty-six plants (Table 1) at various coefficients of inbreeding (F) previously expressed some level of PSC and were chosen to serve as parents. Environmental conditions for flowering followed standard protocols (Scott, 1957), while crossing, seed development, embryo rescue, and germination proceeded using rapid-generation cycling (RGC) (Anderson et al., 1990). Three inflorescences (replications), containing variable numbers of both disc (hermaphroditic) and ray (gynoecious) florets, were self pollinated for each plant each generation. Nomenclature used to describe inbred generations deviated from the standard So to Sn symbol used to designate the original selfed plant and subsequent derived inbreds. The use of S to denote selfed generations would have been confusing for species possessing SI, since S is used for self incompatibility allele designations. Plants were identified by three numbers separated by a period, i.e. 1.63.02, 1.07.01, 1.20.06, etc. The first represented the RGC (1-3), the next designated the cross number within that cycle, and the last was for the plant within each cross, numbered sequentially according to flowering order (Anderson, 1989). Plants that did not possess hermaphroditic disc florets or that failed to shed pollen at anthesis were discarded as male sterile. Male fertility of plants shedding pollen was established using pollen stainability (Owczarzak, 1952), with -> 50% stainable pollen considered the minimum acceptable level. A bulk-outcross pollination was used to establish female fertility, since the occurrence of matched S alleles would be minimized. Thus, an accurate estimate of fertility could be obtained without being confounded with seed set reductions due to SI (An-
158 derson et al., 1990). Bulked pollen was derived from the 66 original parents using a vacuum pollen collector. Three emasculated inflorescences (ray florets only) were pollinated; the number of ovules and seeds were then recorded. Plants failing to set seed following bulk-outcross pollinations were considered female sterile. Female fertility levels were compared with self seed set values using female coefficient of crossability (FCC) calculations (Anderson et al., 1989). FCC values typically range from 0--1 (FCC = 1 when self and outcross values are the same; FCC = 0 when self values are zero; FCC = 0.5 when self values are one-half those of outcrosses). FCC exceeds values of one when self values surpass outcross values. Multiple plant descent, a modification of 'single plant descent' (EI-Nahrawy & Bingham, 1987; Ste-
phens & Bougourd, 1988), was used to advance each inbred generation. In this procedure, each seed from the self pollinations was given an equal chance to contribute gametes to the next generation. If a particular genotype was tolerant of IBD and possessed PSC, it gave rise to a distinct inbred line within the family. Selection was practiced within each inbred population (in every RGC) for male/female fertile plants with PSC. No selection was intentionally performed for horticultural traits. Two stages of the life cycle were chosen to evaluate IBD: seed germination and survivorship (the number of fertile progeny at anthesis). Seed set was recorded but not included in IBD evaluations, since SI and IBD are confounding factors that decrease seed set. The number of embryos surviving
Table 1. Inbreeding coefficients (F values) for 66 chrysanthemum accessions used as parents for generating inbreds for evaluation of inbreeding depression
Family
Parent
Parental coefficient of inbreeding (F)
54-101 - 11
Minnesota Inbred Families 81 -L241-3 0.844
73-55-22
83-172-2 84-161-17,28 84-161-28 x 17 85-117-8 85-312-3 x 6
0.898 0.844 0.898 0.896 0.726
77-AM2
85-303-1,2 81-1-1 83-76-24 83-76-33 x 24 85-341-4,9 85-341-4 x 9
0.812 0.500 0.510 0.520 0.656 0.726
77-AM3-17 79-Z142-2 80-L15-2
0.750 0.875 0.828
84-Al1-1
0.500
77-AM3
82-90-39 'Evening Glow'
'Evening Glow' 84-153-21
0.000 0.625
Family
Parent
Minnesota Inbred Families 'Lindy' 0.000 'Centerpiece' 0.000
'Lindy'
7207
Noninbreds
Parental coefficient of inbreeding (F)
82-119-2,8 83-263-1,6,7 83-263-1 x 6 85-92-51,61 85-92-61 x 51
0.875 0.844 0.898 0.812 0.828
'Golden Star'
0.000
Noninbred Cultivars and Species 'Daisy White' 0.000 'Princess' 0.000 'Yogo' 0.000 'Echo' 0.000
Diploid Species
'Royal Pomp' 'Autumn Glory' 'Glorious' 'Petit Point '1
D. makinoi (= D. ]aponicum)
F1 and F2 Hybrids 86-N711-1 AGSL-1-5,7,10,12 GLSL-2,6-10,12,13 PPSL-1,4-6,9,10-12
0.000
0.000 0.000 0.000 0.250-0.500
1Values range from 0.250-0.500 for this F2 hybrid due to its origin from full-sibbing or selfing F1 hybrid 'Autumn glory' seedlings (see text).
159 to or past the heart stage (visualized as having seed coat coloration 10-15 days post-pollination) in each inflorescence was used as an estimate of self seed set. This served as the base number from which germination data was calculated. Seed germination was estimated using germination data from each set of pollinations per plant. Finally, survivorship (expressed as percent of seed set and percent of germinated seedlings) was determined. Plants counted as survivors had to reach anthesis and be male/female fertile. When rosetted plants appeared, they were sprayed to runoff with 100 ppm GA3. Those failing to flower following GA treatment were subjected to a six-week cold temperature treatment (4.4 ° C) and returned to the glasshouse for flowering. Values for germination and survivorship of consecutive inbred lines were regressed against percent homozygosity (coefficient of inbreeding x 100) for inbred populations surviving into the third generation (Lynch, 1988). A linear model was fitted: Y = b0 + blX, where X is the level of homozygosity. Percent germination and survivorship were correlated (using both percent survivorship calculations). The number and percent of parental lines surviving inbreeding and the average coefficient of inbreeding (F) for lethals and nonlethals were calculated. In addition, germination and survivorship lethals were determined as follows: Germination lethals = Total self seed set - No. of germinated seedlings (nonlethals) Survivorship lethals -- No. of germinated seedlings- No. seedlings reaching anthesis (nonlethals) Lethal and nonlethal F values were either weighted or unweighted, based on the number of individuals at each coefficient of inbreeding.
Results
For the purpose of brevity, data from each of the populations arising from the 66 original parents are not presented. Average values for the inbred fam-
ilies are depicted in Figs. 1 and 2. Sample responses of seed set, germination, and survivorship from seven of the original parents, whose progeny survived into the RGC-3, are included in Table 2. Self and outcross seed set were expressed both as the total number of seeds and as a percentage of the number of ovules present within the inflorescence (Table 2). Percent seed set is critical for the purpose of comparing self and outcross seed set, since ovule numbers differ significantly between inflorescences within plants (Anderson et al., 1989). In addition, the potential number of ovules in a self pollination greatly exceed those for outcross pollination since both disc (hermaphroditic) and ray (gynoecious) florets are used in the former, while only ray florets are used in outcrosses. Percent self seed set ranged from 0.3-27.5% while outcross values ranged from 10.8% to 87.7% for the sample inbred families shown in Table 2. FCC values were primarily between zero and one, most frequently clustered below 0.5. Low FCC values for noninbred parents indicates that a strong SI system and weak expression of PSC were responsible for low percent self seed set. In subsequent inbred generations, the SI system presumably is still strong but it is impossible to separate its effects from those attributable to IBD. In general, percent self seed set was significantly lower than percent outcross seed set. One notable exception to this was 1.52.01 (RGC-2 inbred parent from GLSL-7, Table 2) where the self seed set (27.5%) was approximately twice that of outcrossing (13.6%). Similar comparisons between self and outcross performance for germination and survivorship have been examined elsewhere (Anderson et al., 1992) since they are not the focus of this study. Female fertility levels, as measured by percent outcross seed set, generally declined as inbreeding progressed (Table 2). The inbreds derived from hybrids were frequently exceptions to this, however. For example, 86-N711-1 ('Royal Pomp') and GLSL-7 ('Glorious') had higher outcross seed set percentages in the RGC-3 parents (2.44.01, 2.44.02, 2.66.02) than had been observed in the previous two generations. Despite these gains in female fertility, there was not a corresponding increase in percent self seed set, indicating that SI or
160 Seecl Set
Ill-
N'
]
s4--101-11
]
T'3--88--'~L~
]
7"7--~2
]
?"7--,tAk3 m,qnl~ C...
0 Z
Z4,
2
3
Percent Germination
100.
]
5dlk.101_11
]
T~--~
80,
i
~lU ev*ninS " * w
00.
,tO.
2.0.
0
2
3
Percent Survivorship 100.
80.
._=-
]
it41..to1_11
]
73_88_.~,.~
]
"~'--AM'J
]
77--Nd3
80. ..q U:
Q. 20.
0
1
3
Rapid Generation Cycle
J
Fig. 1. Histograms of average + S.E. self seed set (total no. of seeds from three inflorescences/plant), percent germination, and percent survivorship for three rapid generation cycles (RGC) of hexaploid, inbred chrysanthemums (Dendranthema grandiflora) derived from the University of Minnesota garden chrysanthemum breeding program.
161 Seed
slllo
Set
[] []
i
o
,oo
Percent
D. m W ~ l
;
Germination
I i N~MI N I ~ 0
;
~.
Percent
~N 3
Survivor~hip
'°°1
°'it *] 0
1
]
c~dUv~.t
]
D. rnaid~d
]
Heya PemO
]
P~dt Pe~t
0J 1°
2 Rapld Generation Cycle
3
Fig. 2. Histograms of average ___S.E. self seed set (total no. of seedsfrom three inflorescences/plant), percent germination, and percent survivorship of three rapid generation cycles (RGC) of inbred chrysanthemums derived from diploid Dendranthema makinoi populations, and hexaploid exotic cultivars, and F1 and F2 hybrids ('Royal Pomp', 'Autumn Glory', 'Glorious', and 'Petit Point') of garden and glasshouse chrysanthemums (D. grandiflora).
162 some deleterious recessive genes exposed as a result of inbreeding were being expressed. The lack of correlation between female fertility levels and self seed set would be expected since SI is a prefertilization barrier unrelated to fertility p e r se. Average RGC-1, 2, and 3 family statistics for self
seed set, percent germination, and percent survivorship are depicted graphically in histograms for Minnesota inbreds (Fig. 1) and noninbred cultivars, species, and hybrids (Fig. 2). These histograms summarize the effects of SI and/or IBD on population dynamics. Average seed set values var-
Table 2. Sample progeny sizes for three life cycle stages (seed set, germination, and survivorship) and female coefficient of crossability (FCC) for chrysanthemum families tolerating -> 3 generations of inbreeding (selfing)
Family
Parent
Minnesota Inbreds 73-55-22 84-161-28x 17 1.05.07 2.04.03 77-AM2 81-1-1 1.13.01 2.06.06 2.06.07 2.06.09 77-AM3-17 77-AM3-17 1.19.03 2.12.01 2.12.02 2.12.03 2.12.04 2.12.05 'Lindy' 85-92-61 1.31.01 2.36.01 2.36.02 F1 and F2 Hybrids 'Royal Pomp' 86-N711-1 1.41.23 2.44.01 2.44.02 'Glorious' GLSL-7 1.52.01 2.66.02 'Petit Point' PPSL-12 1.64.01 2.73.01
Progeny cross number
1.05 2.04 1.13 2.06
1.19 2.12
1.31 2.36
1.41 2.44
1.52 2.66 1.64 2.73
Pooled I self seed set
Pooled 1 outcross 2
No. seeds
No. (%) ovules
No. seeds
No. (%) ovules
43 77 3 2 24 4 15 37 19 9 1 6 2 10 10 6 99 6 9
302 483 188 260 240 234 302 289 300 521 342 125 333 343 227 301 394 294 470
(14.2) (15.9) (1.6) (0.8) (10.0) (1.7) (5.0) (12.8) (6.3) (1.7) (0.3) (4.8) (0.6) (2.9) (4.4) (2.0) (25.1) (2.0) (1.9)
97 14 15 179 24 8 22 17 41 21 17 28 27 7 7 43 50 27 10
208 45 35 204 33 74 53 78 50 76 76 67 66 43 54 87 93 85 75
(46.6) (31.1) (42.8) (87.7) (72.7) (10.8) (41.5) (21.8) (82.0) (27.6) (22.4) (41.8) (40.9) (16.3) (13.0) (49.4) (53.8) (31.8) (13.3)
90 4 87 16 5 30 4 124 12 1
1002 323 347 305 395 109 115 741 110 232
(8.9) (1.2) (25.1) (5.2) (1.3) (27.5) (3.5) (16.7) (10.9) (0.4)
70 18 62 122 38 6 12 154 19 3
190 38 71 153 103 44 19 355 40 96
(36.8) (47.4) (87.3) (79.7) (36.9) (13.6) (63.2) (43.4) (47.5) (3.1)
Inbred germination
Inbred survivorship
Total
(%)4
Total
(%)5
0.30 0.51 0.04 0.01 0.14 0.16 0.12 0.59 0.08 0.06 0.01 0.11 0.02 0.18 0.34 0.04 0.47 0.06 0.14
26 11 1 2 20 3 12 19 18 9 1 6 2 6 6 2 3 0 3
(60.5) (14.3) (33.3) (100.0) (83.3) (75.0) (80.0) (51.4) (94.7) (100.0) (100.0) (100.0) (100.0) (60.0) (60.0) (33.3) (3.0) (0.0) (33.3)
8 6 0 2 7 0 2 2 9 5 0 0 1 4 1 1 2 0 1
(30.8) (54.5) (0.0) (100.0) (35.0) (0.0) (16.7) (10.5) (50.0) (55.6) (0.0) (0.0) (50.0) (66.7) (16.7) (50.0) (66.7) (0.0) (33.3)
0.24 0.02 0.29 0.06 0.04 2.02 0.05 0.38 0.23 0.13
61 4 83 16 4 24 0 34 5 0
(72.6) (100.0) (95.4) (100.0) (80.0) (80.0) (0.0) (27.4) (41.7) (0.0)
36 2 37 6 3 1 0 24 1 0
FCC3
seed set
(59.0) (50.0) (44.6) (37.5) (75.0) (4.1) (0.0) (70.6) (20.0) (0.0)
1There were n = 3 inflorescences (reps) per self or outcross pollination. 20utcross seed set was performed using bulked pollen from the 66 original parents. This maximized seed set potential and minimized matched S alleles. 3FCC = 1 - [(outcross - self)/outcross] (Anderson et al., 1989). 4 Expressed as a percentage of total self seed set. s Calculated as a percentage of germinated seedlings.
163 ied across R G C generations, but generally decreased as inbreeding progressed. Values from two families, i.e. 54-101-11 (Fig. 1) and 'Royal Pomp' (Fig. 2), had higher means in the RGC-1 since only one parent represented the family. In subsequent generations this was not the case, since there were multiple parents. Similar trends for Minnesota inbreds and hybrids existed for average performance between groups, with average values clustered within close proximity, and higher percentages for germination than for the other two life cycle stages. IBD was severe in noninbreds in the RGC-2, despite the highest percent germination, demonstrating that seed set and germination were not guarantors of obtaining fertile progeny. Seven of the 66 original parents were completely SI (zero seed set) in the RGC-1 crossing environment (data not shown). As has been previously reported by Ronald & Ascher (1975a, 1975b), PSC was environmentally interactive, since these seven parents had exhibited some degree of PSC in previous years (unpublished data). Expression of PSC in the remaining 59 parents was variable with total self seed set from the three inflorescences ranging from one to 360. In cases where relatively small (n < 10) seed set occurred, percent germination and survivorship were frequently skewed towards 100%, whereas greater seed set resulted in values only as high as 95% and 67%, respectively. Seven advanced Minnesota inbreds, two noninbred cultivars, two D. makinoi, and eight F1 and F2 hybrids had no progeny reaching anthesis at the termination of RGC-1. With all of these parents except one (83-76-33 × 24, n = 21 self seeds) the self seed set values were < 10. Families 82-90-39 and 7207 were completely SI in RGC-1, were excluded from further generations, and data from these two families are not presented in Fig. 1. The small seed set for 'Evening Glow' resulted in no fertile progeny (Fig. 1). Thus, these three families did not contribute progeny to the RGC-2. Average seed set and germination in RGC-1 were highest for the Minnesota inbreds (with higher F values) and could be ranked from highest to lowest as follows: Minnesota inbreds > F1 and F2 hybrids > noninbred cultivars and species. However, percent survivorship in both the
noninbred and hybrid groups exceeded the Minnesota inbred group. In RGC-2, all progeny from the 54-101-11 family were SI (Fig. 1). PSC expression was variable in the remaining families. Higher seed set did not result in large numbers of fertile progeny. For example, parents 1.07.01 (n = 367), 1.20.06 (n = 144), and 1.63.02 (n = 96) had 0-1.4% survivorship. In these three cases, the percent germination was also low (0.5-9%). As with RGC-1,100% percent germination occurred only when the seed set values were small (Table 2). At the termination of RGC-2, all noninbred cultivars, D. makinoi, and 'Autumn Glory' had no surviving progeny. No difference in response to inbreeding was evident between diploid and hexaploid individuals. Noninbreds of both ploidy levels were tolerant of inbreeding only to the RGC-2 germination stage. 'Autumn Glory' had less progeny for all three stages of RGC-2 than any of the other hybrids. However, the Minnesota inbred family with the highest average coefficient of inbreeding (73-55-22) exhibited a lower average percent germination than other Minnesota inbred families or, in some cases, hybrids. Seed set values were variable again in RGC-3, but the range was smaller (n = 1 - 98) than for the previous two generations (RGC-I: n = 1 - 360; RGC-2: n = 1 - 367). A similar trend was evident for percent germination, where sells with low seed set had higher germination percentages (as high as 100%). Survivorship did not depend so much on progeny size as on the particular parent within a family. For example, full-sibs 2.12.04 and 2.12.05 (77-AM3-17 family) had equal self seed set and percent germination values, but differed with respect to survivorship (66.7% and 16.7%, respectively) (Table 2). 'Royal Pomp' inbreds were superior to all others in all three stages of the life cycle (Fig. 2; Table 2). Small seed set values were obtained for 73-55-22 and 'Glorious' RGC-3 parents, as well as low germination or survivorship percentages. These families failed to survive the RGC-3 generation, leaving only three Minnesota inbred and two hybrid families. Regression coefficients (bl) for percent germina-
164
tion and survivorship were negative, except germination for 'Glorious' (Table 3), indicating a reduction for both traits as homozygosity increased. Only 'Lindy' had a negative correlation lower than any of the hybrid-derived inbreds for germination. However, 'Lindy' inbreds were also at a lower percent homozygosity than any of the other Minnesota inbreds. P-values were highly significant for the 77-AM2, 77-AM3 (germination only), and 'Glorious' inbreds, as well as the combined Minnesota and hybrid series. Correlations between percent germination and survivorship (as a percent of seed set) were almost identical for Minnesota inbreds (r= 0.67, P = 0.002) and hybrids (r= 0.67, P = 0.006), but slightly lower for all genotypes (r= 0.66, P-< 0.001). These correlations were higher and more significant when percent germination and survivorship (as a percent of germinated seedlings) were compared (r = 0.95, P-< 0.001). Only six of the original 66 parents produced progeny beyond the third generation. Average weighted or unweighted F values for lethals and nonlethals indicated that more individuals with lower levels of inbreeding were being eliminated at
germination in RGC-1 (Table 4). Although IBD at germination was most severe in RGC-1, comparisons of either numbers or percent of parental lines surviving from germination to anthesis (survivorship) indicated that, for all three RGC generations, IBD expressed at germination was greater than that measured for survivorship. Weighted and unweighted F values were not statistically compared, since they are calculations rather than observations. Most rosetted progeny failed to flower following the GA3 and/or cold temperature treatment. Of those that did reach anthesis, all were either male sterile or SI.
Discussion Although most research conducted on IBD or genetic load in perennials involves only one stage of the life cycle, lethalities in separate stages of the life cycle have been considered independent (Stephens & Bougourd, 1988). Since percent germination and survivorship of inbred chrysanthemums were correlated, regardless of the method of calculating
Table 3. Linear regression coefficients + S.E. for combined inbred chrysanthemum families obtained by fitting percent germination values and percent survivorship values (as a percentage of germinated seedlings) of each inbred line to percent homozygosity (Coefficient of Inbreeding x 100) for the selfed rapid generation cycle ( R G C - 1 , 2 , 3 ) series
Family
Parent(s)
Percent germination
Percent survivorship
b0
bl
S.E.
P
b0
bl
S.E.
P
Minnesota inbreds 73-55-22
84-161-28 x 17
354.16
- 3.43
+ 3.22
0.347
562.93
- 5.77
+ 6.37
1.000
77-AM2
81-1-1
157.41
- 1.22
+ 0.17
0.001
213.34
- 2.24
+ 0.22
0.061
77-AM3
77-AM3-17 391.32
- 3.74
+ 1.31
0.016
135.25
- 1.17
+ 1.30
1.000
79-Z142-2
'Lindy'
82-119-2 85-92-61
Combined F, a n d F2 hybrids 'Royal Pomp' 'Glorious'
86-N711-1
+ 1.80
1.000
149.26
- 1.27
+ 1.71
1.000
- 1.57
+ 0.44
0.001
172.62
- 1.57
+ 0.41
0.001
70.64
- 0.06
+ 0.13
1.000
30.65
- 0.02
+ 0.46
1.000
80.68
0.97
+0.09
<0.001
78.32
-1.18
+0.19
0.003
64.05
- 0.19
+ 0.75
1.000
119.16
- 1.34
+ 0.40
74.72
- 0.67
+ 0.23
0.012
74.42
- 0.83
+ 0.24
0.029 0.004
79.56
- 0.22
+0.16
0.180
69.93
-0.50
+0.16
0.004
PPSL-10 PPSL-12
Combined Total Combined
- 0.02
GLSL-6 GLSL-7
'Petit Point'
45.53 188.96
165 percent survivorship, our study does not support the idea of independence of lethality in various stages of the life cycle. In fact, when percent survivorship and germination are both expressed as percentages of seed set, the two calculations are mathematically related and the calculated r value may derive more significance from this mathematical relationship than a biological one. However, survivorship expressed as a percent of germinated seedlings is the more realistic portrayal of relatedness between germination and survivorship, and this relationship exhibited the higher r value. Lethal and sublethal types were not restricted to the early generations of selfing and surfaced in every inbred family. While the coefficient of inbreeding approached 100% (F = 0.995) for the inbred family 73-55-22, this did not eliminate the expression of post-germination IBD. Results with noninbred materials were similar to those reported for Trifolium (Duncan et al., 1973), Ipomoea (Martin, 1973), and Fagopyrum (Komaki, 1982), where the number of lines that could be maintained after the third generation of selfing was small. Tabulating the number and percent change of parental lines surviving and weighted/unweighted
F values for lethals and nonlethals (Table 4) confirmed the observations (Figs. 1,2) that the first 1-2 generations of inbreeding were the most rigorous in eliminating genotypes with deleterious alleles upon selling noninbred parents. This was due to the expression of lethal or sublethal genes uncovered as a result of the inbreeding process itself and not due to SI expression directly. Because advanced Minnesota inbreds had already survived this screen, a smaller portion of these inbred progenies were lost in RGC-1 and 2. The question arises as to why advanced inbreds often showed a greater (negative) slope in the linear regression model (Table 3) than the hybrid derived inbreds. This is explainable by understanding that in all advanced inbred generations of any species (regardless of ploidy), the variances are large as the F values asymptotically approach one (Lynch, 1988). Linear regression coefficients indicated that there were differences in the rates of IBD between and among inbred families (Table 3). Some P values were highly significant, while others were not. The question might be posed whether quadratic regression would provide a better fit. Quadratic regressions were not performed in this experiment
Table 4. Population dynamics of germination and survivorship used as gauges of inbreeding depression for the 66 initial chrysanthemum parents self pollinated for three rapid generation cycles of inbreeding Rapid Generation Cycle (RGC)
Number (%) of parental lines surviving3
Percent change4
Average coefficients of inbreeding (F) Nonlethals
Lethals 1'2
Weighted F 5
Unweighted F
Weighted F 5
Unweighted F
Germination 1 2 3
49 (74.2%) 23 (34.8%) 7 (10.6%)
- 25.8% - 33.4% - 10.6%
0.806 0.869 0.935
0.696 0.854 0.951
0.635 0.917 0.946
0.677 0.854 0.941
Survivorship 1 2 3
45 (68.2%) 14 (21.2%) 6 (9.1%)
- 6.0% - 13.6% - 1.5%
0.691 0.907 0.924
0.71 0.869 0.949
0.861 0.858 0.93
0.741 0.853 0.951
1Germination lethals = Total selfed seed set - Number of germinated seedlings (Nonlethals). 2Survivorship lethals = Number of germinated seedlings - Number reaching anthesis (Nonlethals). 3There were 66 parents at the beginning of the experiment. 4Percent change was calculated by subtracting the percentage of parental lines surviving from the previous stage's surviving percentage (either germination or survivorship). The initial percentage of parental lines (66) = 100%. 5Weighted F indicates that each F value was weighted based on the number of individuals at that level of inbreeding.
166 since only three groups of data points (generations) were available for analysis. This is the minimum number necessary for linear regression; quadratic analysis would require at least four points. Regression of three points that are not quite linear would always fit a quadratic equation better. A fourth generation would enable one to perform both linear and quadratic regression coefficients. Since some level of dominance is necessary for the expression of IBD (Good & Hallauer, 1977; Bingham, 1979), a decrease in response as homozygosity increased would be expected. The negative regression coefficients (bl) indicated that, indeed, IBD in some chrysanthemum families could be explained by a genetic model based on the cumulative effects of loci with dominance. Although dominance was important, however, epistatic effects (confounded with different environments in this study) may also play a role in IBD, since deviations from linearity would be an indicator of epistasis (Hallauer & Sears, 1973; Good & Hallauer, 1977). The number of loci affecting the traits could also have been a factor for determining the rate of decline with increased homozygosity. Since significant changes occurred after three generations of selfing Minnesota inbreds, it is conceivable that more than two or three major loci were involved (Hallauer & Sears, 1973), especially since this is a hexaploid species. Linear regression of germination (Y = 79.560.223X, r 2= 0.17, P = 0.18) and survivorship (Y = 69.93- 0.501X, r 2= 0.49, P = 0.004) for the total combined values indicated that no detectable germination variation in IBD was accounted for by F. A concern in this experiment was the relevance of the coefficient of inbreeding (F) for an allohexaploid. In diploid organisms, F is the measure of the probability that two alleles are identical by descent from a common ancestor (Wright, 1922). Wright assumed that, on average, all loci approached homozygosity at the same rate, independent of gene frequencies, and that, theoretically, an individual with F = 1 would be completely homozygous (Boucher, 1988). Recent studies involving electrophoretic variability of loci suggest that this hypothesis does not necessarily apply to unselected loci (Smith et al., 1985; Smith & Smith,
1987; Alan Durrant, personal communication, 1988). Genetic variability frequently exists at loci not under selection pressure. When F is applied to polyploids, the situation becomes far more complex. The only option that has been proposed is alteration of the F equation to account for meiotic pairing configurations: homologous for allopolyploids and homoeologous for autopolyploids. In chrysanthemums, homologous pairing is the norm and homoeologous pairing is rare. Thus, the diploid equation was applied to calculate F. Regardless of the equation, however, the use of F to depict homozygosity in polyploids becomes tenuous for several other reasons. First, recessive alleles can occur in many more combinations in polyploids (nulliplex, simplex, duplex, triplex, etc.), while still being expressed only in the homozygous recessive (nulliplex) state. Since the proportion of this class is reduced (relative to diploids) as the level of ploidy increases, it should take a greater number of generations for IBD caused by nulliplex configurations to surface. These theoretical expectations do not hold true for chrysanthemums, however, as IBD was significant in progeny derived from noninbred parents in RGC-1 and RGC-2. Second, the diploidizing mechanism in allopolyploids (Haufler, 1987) may not have complete penetrance, causing less than 100% homologous pairing. Homoeologous pairing could then occur with differing frequencies between plants and generations, creating bridges and acentric fragments at meiosis and, subsequently, varying degrees of male and female sterility. Sterilities resulting from homoeologous pairing, although independent of F, would be confounded with IBD. Third, polyploids could be homozygous for different alleles at duplicated loci when F = 1. This fixed heterozygosity should minimize the effects of IBD (Stebbins, 1957; Libby et al., 1981; Haufler & Soltis, 1986). However, homoeologous pairing at any generation of inbreeding could result in homozygosity across duplicated loci and the reappearance of IBD, regardless of F. Finally, as much as 50% of the duplicated genes are silenced in ancient polyploids (Ferris & Whitt, 1977). Should anything in the process of inbreeding
167 alter this silencing, waves of IBD would appear, independent of F, as deleterious-recessive alleles in previously-silenced genomes were expressed. Thus, it is evident that no F equation proposed to date can adequately encompass all of the genetic mechanisms that could be in operation in allohexaploid chrysanthemums. We chose to apply the formula for disomic inheritance, given the preponderance of bivalents. While this and any other F formulae may be inadequate descriptors of homozygosity, there are no other alternatives at present. Comparable or superior performance of the full sib (84-161 - 28 x 17) and hybrid derived inbreds (86-N711-1, GLSL-6, GLSL-7, PPSL-10, PPSL-12) in relation to the advanced Minnesota inbreds merits attention. Chance alone could deem that the selected crosses would have been good performers, since the hybrid and full-sib-inbred sample size was small. However, since other full-sib (85-312-3 x 6; 83-76-33 x 24; 85-341-4 × 9; 83-263-1 × 6; 85-926 1 x 5) and hybrid (AGSL-1-5, 7, 10 and 12; GLSL-2, 8-10, 12, and 13; PPSL-1, 4--6, and 9) inbreds were not tolerant of three generations of inbreeding due to the singular or joint expression of SI and IBD, the specific crosses noted above may be favorable recombinants. Because full-sib mating has been used in the chrysanthemum breeding program to circumvent the occurrence of SI, it is not surprising that only one out of six full-sibderived inbreds was also tolerant of IBD and possessed PSC. Since inbreeding and the rate of IBD progresses much slower with full-sib mating than with selfing, the inbreds derived from 84-161-28 x 17 would possibly succumb to IBD at some point in future generations. Nonetheless, this demonstrated that IBD could be minimized in some inbreds derived from full-sib mating. Bailey (1971) proposed the term recombinant inbred strains for those inbreds derived 'from the cross of two unrelated but highly inbred progenitor strains and which have been maintained independently under a regiment of strict inbreeding since the F2 generation'. Some of the hybrids used in this experiment would classify as recombinant inbred strains. In inbred families where IBD or SI are present (all cases presented here), it would be advantageous to carefully juxtapose inbreeding and
outcrossing, thereby realizing greater gains than would otherwise be possible (Campbell, 1988). The superior performance of the recombinant inbreds does suggest that the method of inbreeding chrysanthemums and perhaps other polyploid, outcrossing species should not follow the standard regiment of selling or full-sib mating to achieve homozygosity.
References Allard, R.W., S.K. Jain & P.L. Workman, 1968. The genetics of inbreeding populations. Advances in Genetics 14: 55-131. Anderson, N.O., 1987. Reclassifications of the genus Chrysanthemum L. HortScience 22(2): 313. Anderson, N.O., 1989. Rapid generation cycling and inbreeding depression in chrysanthemums. PhD Thesis. University of Minnesota. Anderson, N.O., B.E. Liedi, P.D. Ascher, R.E. Widmer & S.L. Desborough, 1988. Evaluating self incompatibility in Chrysanthemum. The influence of ovule number. Sexual Plant Reproduction 1(3): 173-181. Anderson, N.O., B.E. Liedl, P.D. Ascher & S.L. Desborough, 1989. Distinguishing between self-incompatibility and other reproductive barriers in plants using male (MCC) and female (FCC) coefficient of crossability. Sexual Plant Reproduction 2: 116-126. Anderson, N.O., P.D. Ascher, R.E. Widmer & J.J. Luby, 1990. Rapid generation cyclingof chrysanthemum, Dendranthema grandiflora Tzvelv. (Chrysanthemum morifolium Ramat.). Journal of the American Society for Horticultural Science 115(2): 329-336. Anderson, N.O., P.D. Ascher & R.E. Widmer, 1992. Lethal equivalents and genetic load. Plant Breeding Reviews (In Press). Ascher, P.D., 1976. Self incompatibility systems in floricultural crops. Acta Horticulturae 63: 205--215. Bailey, D.W., 1971. Recombinant-inbred strains: An aid to finding identity, linkage, and function of histocompatibility and other genes. Transplantation 11(3): 325-327. Bingham, E.T., 1979. Maximizing heterozygosity in autopolyploids, p. 471-489. In: Lewis (Ed). Polyploidy: Biological relevances. Plenum, New York. Boucher, W., 1988. Calculation of the inbreeding coefficient. Journal of Mathematical Biology 26: 57~a4. Brewer, J.G., 1968. Flowering and seed setting in pyrethrum (Chrysanthemum cinerariaefolium Vis.). A review. Pyrethrum Post 9: 18-21. Campbell, R.B., 1988. Mating structure and the cost of deleterious mutation: 1. Postponing inbreeding. Journal of Heredity 79: 179-183. Crook, C.B., 1942. Genetic studies of chrysanthemum. MS Thesis. Kansas State University.
168 Darwin, C.R., 1876. The effect of cross- and self-fertilization in the vegetable kingdom. John Murray, London. Dowrick, G.J. & A.S. EI-Bayoumi, 1966. The origin of new forms of the garden chrysanthemum. Euphytica 15: 32-38. Drewlow, L.W., P.D. Ascher & R.E. Widmer, 1973. Genetic studies of self incompatibility in the garden chrysanthemum, Chrysanthemum morifolium Ramat. Theoretical and Applied Genetics 43(1): 1-5. Duncan, J.F., N.K. Anderson & N.L. Taylor, 1973. Effect of inbreeding on pseudo-self compatibility in red clover (Trifolium pratense L.). Euphytica 22: 535-542. E1-Nahrawy, M.A. & E.T. Bingham, 1987. Inbreeding depression as a measure of improvement in alfalfa (Medicago sativa L). Agronomy Abstracts (Abstr.). Evans, G.M. & A.J. Macefield, 1972. The suppression of homoeologous pairing by B-chromosomes in a Lolium species hybrid. Nature 236: 110--111. Ferris, S.D. & G.S. Whitt, 1977. Loss of duplicate gene expression after polypioidisation. Nature 265: 258-260. Fowler, D.P., 1965a. Effects of inbreeding in red pine, Pinus resinosa Ait. II. Pollination studies. Silvae Genetica 14(1): 12-23. Fowler, D.P., 1965b. Effects of inbreeding in red pine, Pinus resinosa Ait. IV. Comparison with other Northeastern Pinus species. Silvae Genetica 14: 76-81. Good, R.L. & A.R. Hallauer, 1977. Inbreeding depression in maize by selfing and full-sibbing. Crop Science 17: 935-940. Hallauer, A.R. & J.H. Sears, 1973. Changes in quantitative traits associated with inbreeding in a synthetic variety of maize. Crop Science 13: 327-330. Haufler, C.H. & D.E. Soltis, 1986. Genetic evidence suggests that homosporous ferns with high chromosome numbers are diploid. Proceedings of the American Academy of Science 83: 4389-4393. Haufler, C.H., 1987. Electrophoresis is modifying our concepts of evolution in homosporous pteridophytes. American Journal of Botany 74(6): 953-966. Johannsen, W., 1926. Elemente der exakten Erblichkeitslehre. Jena, Fischer. 3rd. ed. Kawase, K. & Y. Tsukamoto, 1966. Studies on the breeding of chrysanthemum. 10. On the inbreeding, p. 289-290. Japanese Society for Horticultural Science. Spring Meeting (Abstr.). Kawase, K. & Y. Tsukamoto, 1977. Studies on breeding in chrysanthemum, Chrysanthemum morifolium Ramat. I. Selffertility. Engei gakkai zasshi 46(1): 101-112. Kempthorne, O., 1957. An introduction to genetic statistics. John Wiley, New York. Komaki, M.K., 1982. Inbreeding depression and concealed deleterious mutations in buckwheat populations, Fagopyrum esculentum. Japanese Journal of Genetics 57: 361-370. Libby, W.J., B.G. McCutchan & C.I. Millar, 1981. Inbreeding depression in selfs of redwood. Silvae Genetica 30: 15-25. Lynch, M., 1988. Design and analysis of experiments on random drift and inbreeding depression. Genetics 120: 791-807. Marshall, H.H., 1973. Problems in breeding seed lines of garden chrysanthemums: self-sterility and color, p. 97-100. In: West-
ern Canada. Society for Horticulture. Banff, Alberta. Report of Proceedings, V. Martin, F.W., 1973. Fertility, incompatibility, and morphological abnormalities of sweet potato inbreds. Incompatibility Newsletter 2: 29-31. Mulford, F.L., 1937. Results of selfing twenty-four early blooming chrysanthemums. Proc. of the American Society for Horticultural Science 35: 818-821. Niwa, T., 1931. Pollination and self-incompatibilityin chrysanthemum. (Outlines). Japanese Association for the Advancement of Science. Report 6: 479-487. Owczarzak, A., 1952. A rapid method for mounting pollen grains. Stain Technology 27: 249. Rajhathy, T., 1971. The allopolyploid model in Avena. In: G. Kimber & G.P. Redei (Eds). Stadler Genetics Symposium, Vol. 3: Missouri Agricultural Experiment Station, Columbia. Ronald, W.G. & P.D. Ascher, 1975a. Effects of high temperature treatments on seed yield and self incompatibility in chrysanthemum. Euphytica 24(2): 317-322. Ronald, W.G. & P.D. Ascher, 1975b. Self compatibility in garden chrysanthemums: Occurrence, inheritance, and breeding potential. Theoretical and Applied Genetics 46(1): 45-54. Schemske, D.W., 1983. Breeding system and habitat effects on fitness components in three neotropical Costus (Zingiberaceae). Evolution 37(3): 523-539. Schemske, D.W. & R. Lande, 1985. The evolution of selffertilization and inbreeding depression in plants. II. Empirical observations. Evolution 39(1): 41-52. Scott, E.L., 1957. The breeders handbook. Handbook No. 4. National Chrysanthemum Society. Bogota, New Jersey, Dancey Printing. Smith, J.S.C., M.M. Goodman & C.W. Stuber, 1985. Genetic variability within U.S. maize germplasm. II. Widely-used inbred lines 1970 to 1979. Crop Science 25: 681-685. Smith, J.S.C. & O.S. Smith, 1987. Associations among inbred lines of maize using electrophoretic, chromatographic, & pedigree data. I. Multivariate and cluster analysis of data from Lancaster Sure Crop derived lines. Theoretical and Applied Genetics 73: 654--664. Soltis, D.E., 1986. Isozyme number and enzyme compartmentalization in Equisetum. American Journal of Botany 73: 908-913. Stebbins, G.L., 1957. Self-fertilization and population variability in the higher plants. American Naturalist 91: 337-354. Stephens, J.P. & S.M. Bougourd, 1988. Inbreeding depression and the outcrossing rate in natural populations of Allium schoenoprasum L. (wild chives). Heredity 60: 257-261. Stephens, L.C., P.D. Ascher & R.E. Widmer, 1983. Interaction among sporophytic S loci in self-incompatible garden chrysanthemums. Euphytica 33: 623-631. Tanaka, R., 1952. Vergleichende Untersuchungen fiber die Fertilit~itbei Selbstbest~iubung in der Polyploide von Chrysanthemum. Japanese Journal of Genetics 27(1-2): 1-2. Thorpe, H.C., 1940. Pyrethrum breeding. East African Agricultural and Forestry Journal 5: 364-368.
169 Tsukamoto, Y., S. Kosuge & S. Kumo, 1964. Studies on the breeding of chrysanthemums. 6. Results of crossing and selling. p. 36. Japanese Society for Horticultural Science. Autumn Meeting (Abstr.). Watanabe, K., 1977a. The control of diploid-like meiosis in polyploid taxa of Chrysanthemum (Compositae). Japanese Journal of Genetics 52(2): 125-131. Watanabe, K., 1977b. Successful ovary culture and production
of F1 hybrids and androgenic haploids in Japanese Chrysanthemum species. Journal of Heredity 68(5): 317-320. Wright, S., 1922. Fisher's theory of dominance. American Naturalist 63: 247. Wu, L., 1963. A study of chrysanthemum breeding in the southwest. MS Thesis. Texas Technical College. Zagorski, J.S., P.D. Ascher & R.E. Widmer, 1983. Multigenic self incompatibility in hexaploid Chrysanthemum. Euphytica 32: 1-7.