Evolutionary Ecology (2006) 20: 173–191 DOI: 10.1007/s10682-005-5748-5
Springer 2006
Research article
Maintenance of inbreeding depression in a highly self-fertilizing tree, Magnolia obovata Thunb KIYOSHI ISHIDA Kansai Research Center, Forestry and Forest Products Research Institute, Momoyama, Fushimi, 612-0855, Kyoto, Japan (tel.: +81-611-1201; fax: +81-611-1207; e-mail: ishidak@affrc.go.jp)
Received 2 April 2005; Accepted 29 November 2005 Co-ordinating editor: S.-M. Chang
Abstract. Inbreeding depression is a major selective force that maintains outcrossing in flowering plants. If the long life and large mature size of trees cause high inbreeding depression via mitotic mutations and half-sib competition, these characteristics may increase inbreeding depression sufficiently to maintain traits that facilitate outcrossing even with high primary selfing rates (proportion of selfed ovules). Here, I report the maintenance of inbreeding depression in a population of a tree (Magnolia obovata Thunb.) with primary selfing rates greater than 0.8 resulting from geitonogamy. The progenies exhibited inbreeding depression for germination, seedling survival, and seedling mass (d = 0.29–0.38), but no significant difference between crossing type in seedling height. Cumulative inbreeding depression for early survival (from zygote to 2-year-old stage) estimated from these results and from prior data on embryonic survival was high (de = 0.91). The fixation index at maturity based on six allozyme loci was low (Fis = 0.08), indicating that significant inbreeding depression for late survival results in a low level of inbreeding with respect to gene transmission to the next generation. From these results, I estimated that inbreeding depression for late and lifetime survival equaled 0.69 and 0.97, respectively. These results suggest that M. obovata trees maintain high inbreeding depression at both early and late life stages, resulting in a low level of inbreeding despite a high primary selfing rate. The high inbreeding depression can be explained by previous theories and is consistent with the predicted maintenance of inbreeding depression in highly self-fertilizing tree populations. The inbreeding load due to the high primary selfing rate represents a cost of this tree’s pollination system for outcrossing, which is based on automimicry and mass flowering. Key words: fixation index, germination, life history, mating system, outcrossing, pollination system, seedling growth, seedling survival, selective interference
Introduction Inbreeding depression is a major selective force that maintains outcrossing in flowering plants (Lloyd, 1979; Uenoyama et al., 1993). Inbreeding depression is usually measured by the d value as d = 1 ) Ws/Wo, where Ws and Wo are the mean fitness of selfed and outcrossed progeny, respectively (Johnston and Schoen, 1994). Theoretical studies have predicted that a selfer, which reproduces by both selfing and outcrossing, takes advantage over an
174 outcrosser in gene transmission to the next generation, but that this selective advantage decreases as the d value increases (e.g., Barrett and Harder, 1996). On the other hand, evolution of the magnitude of inbreeding depression would depend on the level of inbreeding; inbreeding depression is primarily caused by deleterious recessive mutations (Lynch and Walsh, 1998; Charlesworth and Charlesworth, 1999; Roff, 2002), and previous studies have suggested that successive generations of selfing purge the deleterious recessive mutations via selection, leading to decreased inbreeding depression (Charlesworth et al., 1990; Barrett and Charlesworth, 1991; Crnokrak and Barrett, 2002). Based on this genetic mechanisms, theoretical studies have suggested that complete outcrossing is evolutionarily stable when the d value is higher than a threshold value (0.5 in the simple case), but that complete selfing should evolve for lower values of this parameter (Lande and Schemske, 1985; Uyenoyama et al., 1993). Although such simple models predict disruptive selection on the selfing rate at zygote stage, intermediate selfing rates at zygote stage may be favored, if additional factors such as mode of self-fertilization (Lloyd, 1979) or pollen discounting (Holsinger, 1991) are incorporated into the model. Here, I denote the selfing rate at zygote stage and that at maturity as ‘primary selfing rate’ and ‘secondary selfing rate’, respectively. These predictions for the evolutionary relationship between inbreeding depression and mating systems have promoted empirical studies on plant inbreeding depression (Husband and Schemske, 1996; Byers and Waller, 1999; Cheptou and Schoen, 2003). Despite progress in understanding the coevolution of inbreeding depression and plant mating systems, few studies have focused on the evolutionary relationship between mating system and life history strategy with respect to inbreeding depression. In a review of 217 species, Barrett et al. (1996) found that the distribution of selfing rates measured in seeds or seedlings (rm) for annual and perennial herbaceous species are bimodal, with low (rm < 0.2) and high (rm > 0.8) modes, whereas that for woody perennial species has an L shape, with only a low mode (rm < 0.2). Barrett and his colleagues suggested that the variation in mating system among plants with different life history strategies may arise from differential inbreeding depression that results from different life span (Barrett and Eckert, 1990; Barrett et al., 1996), since previous studies have suggested that the rate of mutation to produce deleterious recessive alleles per generation increases with a plant’s life span because of mitotic mutations (Klekowski and Godfrey, 1989; Klekowski, 1998). In fact, predominantly outcrossing conifers mostly exhibit higher inbreeding depression (lethal equivalents) than annual plants for embryonic survival (Lynch and Walsh, 1998). The theoretical study concerning mutation–selection balance in perennials (Morgan, 2001) argued the significant role of mitotic mutation in the main-
175 tenance of high inbreeding depression of perennials; the model predicts that, in the absence of substantial mitotic mutation, perennials have lower inbreeding depression than annuals with the same mating system. In contrast, if we assume large contributions of recessive mitotic mutations in perennials, as suggested by Klekowski (1998), the model predicts that long-lived perennials, such as trees, will exhibit high inbreeding depression compared with annuals, and that the decrease in their inbreeding depression with increasing primary selfing rate (the proportion of selfed ovules) is so small that d is greater than the theoretical threshold (0.5) even if the primary selfing rate is greater than 0.8. Accordingly, in trees with long life span, the mitotic mutation may cause high inbreeding depression sufficiently to maintain traits that facilitate outcrossing even under high primary selfing rates. If inbreeding depression is high in tree populations, as predicted in the theoretical study, low levels of inbreeding (i.e., low level of inbreeding coefficient or secondary selfing rate) due to high inbreeding depression would also contribute to the maintenance of inbreeding depression under high primary selfing rates, since the effectiveness of selection at purging deleterious recessive mutations is determined by the level of inbreeding (i.e., selective interference among loci; Lande et al., 1994). In addition, severe half-sib competition between selfed and outcrossed progenies may facilitate high values of inbreeding depression for late survival (survival rate after germination) in trees because of their large mature size and long prematurity stage (Lloyd, 1980). Although genetic drift due to small population size might decrease the predicted contributions of mitotic mutations, selective interference, and sib competition by fixing or purging deleterious recessive mutations (Hedrick, 2002; Glemin, 2003), this effect is small unless populations are isolated for successive generations (Glemin et al., 2003). Following these predictions, the tree’s life history characteristics (long life span and large mature size) will maintain high inbreeding depression, resulting in the predominance of outcrossing in tree populations. Testing these predictions requires information on inbreeding depression in highly self-fertilizing tree populations and on its consequences for the level of inbreeding, but few studies have focused on inbreeding depression or the level of inbreeding in such species (Hardner and Potts, 1997; Potts and Wiltshire, 1997; Ishida et al., 2003) and cumulative inbreeding depression (including both early and late life stages) has not been studied in such populations. In the present study, I refer to primary selfing rates higher than 0.6 as ‘high’, since models of inbreeding depression in perennials (e.g., Morgan, 2001) suggest that if mitotic mutations contribute effectively to inbreeding depression, then the difference in the d value among life history strategies would be pronounced at high primary selfing rates (>0.6).
176 In this paper, I report the maintenance of inbreeding depression and the level of inbreeding in the highly self-fertilizing tree species Magnolia obovata Thunb. (M. hypoleuca Sieb. et Zucc., Ueda, 1986). Previous pollination experiments and allozyme analysis of three large populations found relatively stable primary selfing rates (primary selfing rates estimated from embryonic inbreeding depression and selfing rate at the seed stage were 0.82–0.86, respectively; Ishida et al., 2003; K. Ishida, Kansai Research Center, unpublished data). These primary selfing rates are the highest levels reported among self-fertilizing tree species (Husband and Schemske, 1996; Kennington and James, 1997; Potts and Wiltshire, 1997). Despite these high primary selfing rates, M. obovata exhibits a high level of inbreeding depression for embryonic survival (d = 0.76 and 0.78 for two populations; Ishida et al., 2003; K. Ishida, Kansai Research Center, unpublished data). In the present study, I addressed the following questions: Do the progenies of M. obovata exhibit high levels of inbreeding depression for both early and late traits? Does cumulative inbreeding depression in this species result in the low level of inbreeding (i.e., a low inbreeding coefficient and a low selfing rate at maturity)? To answer these questions, I conducted a germination test, nursery experiments, and allozyme analysis to measure levels of inbreeding depression for early survival (germination rate, seedling survival rate), seedling growth (height and mass of 2-year-old seedlings), and the fixation index at maturity in a highly self-fertilizing M. obovata population, whose primary selfing rate and inbreeding depression for embryonic survival have been previously reported (Ishida et al., 2003). In addition, I indirectly estimated the inbreeding depression for lifetime survival (cumulative survival from zygote to maturity) and the secondary selfing rate at maturity from the estimated fixation index and prior data on primary selfing rates. Based on these results, I discuss mechanisms for the maintenance of inbreeding depression under a high primary selfing rate and the ultimate factors responsible for the high primary selfing rate in the species. Materials and methods Study species and site Magnolia obovata is a common deciduous tree native to temperate forests in Japan, and it grows to ca. 20 m (Sano, 1997). Beetles, hover-flies, bees, and thrips visit the flowers (Kikuzawa and Mizui, 1990; Yasukawa et al., 1992; K. Ishida, Kansai Research Center, unpublished data). The flowers are hermaphroditic and protogynous, but their asynchronous flowering within a tree causes high self-fertilization as a result of frequent geitonogamy (Ishida et al., 2003; Isagi et al., 2004). The seeds are a favorite food of birds and are mainly
177 dispersed by these birds. Microsatellite analysis conducted in a natural population revealed that the pollen-flow distance ranged from 3.2 to 540 m and that the average distance between parents and seedlings was 264 m (Isagi et al., 2000), indicating a large effective population size for the species. Seeds for the germination test and nursery experiment and bud materials for allozyme analysis were collected from trees in a deciduous broadleaved forest in Hitsujigaoka (lat. 4300¢ N, long. 14124¢ E, elevation 150 m), Sapporo, Hokkaido, northern Japan, where M. obovata has continuously maintained a large population. Dominant tree species in the forest are Quercus crispula, Tilia japonica, Acer mono, and Betula platyphylla. Previous pollination experiments and allozyme analysis conducted in this population reported that the primary selfing rate and level of inbreeding depression for embryonic survival were 0.86 and 0.76 ± 0.08 (population level value ± jackknife standard error), respectively (Ishida et al., 2003). The germination test and the nursery experiment were conducted at the Forestry and Forest Products Research Institute (FFPRI) in Hitsujigaoka. Germination test and nursery experiment To measure inbreeding depression for early survival (germination and seedling survival rates) and seedling growth (seedling height and mass), I conducted a germination test and nursery experiment using selfed and outcrossed seeds. Prior data on embryonic survival in the study population (Ishida et al., 2003) was included in the present study to estimate cumulative inbreeding depression for early survival. To obtain the selfed and outcrossed seeds, I conducted manual self- and cross-pollinations on five trees randomly chosen from the population in 1996 and 1997 (for a detailed description of the methods, see the prior study; Ishida et al., 2003). The number of selfed and outcrossed seeds obtained per maternal tree ranged from 113 to 370 and 161 to 766, respectively. The seeds were stored over winter on moist filter paper at 5 C to promote synchronous germination. In early May, the seeds were placed on moist filter paper in Petri dishes in an incubator (12 h day/night; 25 C day/5 C night) following the method of Katsuta (1998). The seeds germinated from late May to early July. As M. obovata seeds of the same cohort can germinate over a 2-year period (Satoo, 1992), I again stored ungerminated seeds at 5 C in early August and examined their germination from the following May to July using the same procedure as in the first year. I calculated a germination rate for each subfamily (here, selfed or outcrossed progenies sharing a maternal tree): germination rate ¼ðnumber of seeds that germinated during 2 yearsÞ=ðinitial number of seedsÞ
178 I transplanted seeds that germinated during the first year into 15-l pots in the nursery of FFPRI. The initial number of selfed and outcrossed seedlings per subfamily ranged from 4 to 67 and 4 to 212 in 1996 and 1997, respectively. Either 4 or 3 seedlings were planted in each pot at a 15-cm spacing. In early September of the next year, when the 2-year-old seedlings had ceased their height growth, I measured the seedling height and calculated the seedling survival rate after germination for each subfamily. At the time of the measurement, shading by neighboring seedlings was negligible. The seedlings were then harvested and their total dry mass was determined after drying at 80 C for 72 h. Based on the germination rate, seedling survival rate, and prior data on embryonic survival rate (number of seeds/initial number of ovules; Ishida et al., 2003), I calculated the cumulative survival rate (mi, survival rate from zygote to 2-year-old stage) for each subfamily as mi ¼ me mg ms whereme, mg, and ms are the embryonic survival rate, germination rate, and seedling survival rate, respectively. I examined the effects of crossing type (selfed vs. outcrossed) and family (effects of the maternal tree) on the three components of early survival (germination rate, seedling survival rate, and cumulative survival rate) using two-way ANOVA without replication, since replicates could not be obtained for each subfamily. These survival rates were arcsine-transformed prior to analyses to meet the assumption of normality of residuals in ANOVA. I examined the effects of crossing type, family, and interaction between crossing type and family on the two seedling growth parameters (height and mass) using a two-way mixed-model ANOVA with family as a random factor and crossing type as a fixed factor. I used Type III sums of squares to test the significance of these three effects. If the analysis revealed a significant effect of crossing type, then the result indicated that the overall mean for the outcrossed seedlings differed from the overall mean for the selfed seedlings; in other words, the population exhibited significant inbreeding depression. I used the SPSS 9.0 software (SPSS, Chicago, Illinois, USA) for these analyses. The magnitude of the inbreeding depression (d) for each parameter was calculated by d = 1 ) (ws/wo), where ws and wo are the population means for selfed and outcrossed progeny, respectively (Johnston and Schoen, 1994). The standard error of the d value was calculated using a jackknife resampling method (Manly, 1997) for individual maternal trees. Electrophoretic analysis I calculated heterozygosity and the fixation index at maturity (Fis; Wright, 1969) for allozyme loci using winter buds from the M. obovata population. Bud
179 materials were collected from 60 randomly chosen mature trees (diameter at breast height > 10 cm) in an area of ca. 4 ha (ca. 200 200 m) at the study site. Bud materials were stored at 5 C before electrophoresis. To prepare the samples, leaf primordia from the buds were ground in 500 ll of cold extraction buffer (modified based on the method of Shiraishi, 1988). I analyzed the following six enzymes: aspartate aminotransferase (AAT; EC 2.6.1.1), alcohol dehydrogenase (ADH; EC 1.1.1.1), diaphorase (DIA; EC 1.6.*.*), leucine amino-peptidase (LAP; EC 3.4.11.1), menadione reductase (MNR; EC 1.6.99.2), and 6-phosphogluconate dehydrogenase (6PGD; EC 1.1.1.44). I used polyacrylamide vertical-slab gels according to the methods of Davis (1964) and Orstein (1964), and carried out electrophoresis at 4 C with a current of 12.3 mA/cm2 for 150 min. Genetic interpretations of the resulting banding patterns were inferred from segregation patterns with reference to typical subunit structures (Gottlieb, 1982; Crawford, 1983). Six polymorphic loci were used for the heterozygosity and fixation index at maturity (Aat-1, Aat-2, Adh, Dia, Lap, 6Pgd). Indirect estimation of the levels of inbreeding and inbreeding depression I calculated the fixation index at maturity (Fis) using the FSTAT program of Goudet (1995). The Fis value measures the deficiency in heterozygosity compared with that of a Hardy–Weinberg equilibrium, and is equivalent to the inbreeding coefficient unless there is spatial genetic structure (Nei, 1987). I calculated the 95% bootstrap confidence interval for the Fis value based on bootstrapping over loci. I then tested the Fis value for deviation from a Hardy–Weinberg equilibrium using a 5000-example randomization (i.e., the Hardy–Weinberg equilibrium was generated by randomizing alleles among individuals within a population). To estimate inbreeding depression for late survival (dl), I indirectly estimated the magnitude of inbreeding depression for lifetime survival (dw) using the equation of Ritland (1990). This equation provides an estimate of the relative survival of selfed progeny (wr; wr = 1 ) dw) from the primary selfing rate (r) and inbreeding coefficient (F) as wr = 2(1 ) r)F/[r(1 ) F)], based on two assumptions: constant inbreeding coefficients across the generations and an absence of biparental inbreeding. The bias in the estimate resulting from violation of the former assumption is generally small (Lynch and Walsh, 1998), and that resulting from the latter assumption would be also small in the present study population, since previous allozyme analysis indicated that the level of biparental inbreeding is low in this population: the mean [tm ) ts] value = 0.05 (Ishida et al., 2003), where tm and ts represent the multilocus and mean singlelocus outcrossing rates at the seed stage, respectively. Assuming F = Fis (i.e., neither biparental inbreeding nor population genetic structuring is important),
180 I calculated the dw value by substituting the estimate of Fis for F and the prior data for r (r = 0.86; Ishida et al., 2003) in the equation of Ritland. I then calculated the inbreeding depression for late survival (dl) based on this indirect estimate of dw and cumulative inbreeding depression for early survival (de) value as dw = 1 ) (1 ) de)(1 ) dl). The 95% bootstrap confidence intervals for the dw value and dl value were calculated from 1000 bootstrap samples for the tm values of the population (resampling of seeds within each family; Ishida et al., 2003) by the method of Manly (1997). In addition, I calculated the secondary selfing rate (r*) from the estimate of dw and the prior data on r using the equation of Lande et al. (1994) as follows: r ¼ rð1 dW Þ=ð1 rdW Þ:
ð1Þ
Results Inbreeding depression for early survival and seedling growth Measurements for the germination rate and seedling survival rate from selfed progenies were both 62% of the corresponding values for outcrossed progenies, and d values for both components were 0.38 (Table 1). The cumulative survival rate from fertilization to the 2-year-old stage for selfed progenies was only 9% of the corresponding value for outcrossed progenies, and its d value (de) was 0.91. The ANOVA results indicated that the effect of crossing type was significant for these three parameters, indicating that significant inbreeding depression was exhibited during early survival (Table 2). The effect of family was also significant for the seedling survival rate and cumulative survival rate. The d values for seedling height and seedling mass were 0.19 and 0.29, respectively (Table 1). The ANOVA results indicated that the effect of crossing type was significant for seedling mass, but not for seedling height (Table 3). For these two parameters, the effects of family were also significant, but the interactions between crossing type and family were not significant. Level of inbreeding and indirect estimates of inbreeding depression The mean observed and mean expected heterozygosity for the six allozyme loci were 0.381 and 0.414, respectively (Table 4). The fixation index (Fis) at maturity for each locus and for all loci was )0.103 to 0.307 and 0.080, respectively (Table 4). The Fis value for all loci was marginally significantly greater than zero (p = 0.056; calculated by the randomization test), whereas no Fis value for an individual locus differed significantly from zero (p > 0.05/ 6 = 0.083; 0.083 is the value for a 5% nominal level adjusted by the number of
181 Table 1. Mean and inbreeding depression (d value ± jackknife standard error) for early survival (germination rate, seedling survival rate, and cumulative survival rate) and seedling growth (height and mass) of Magnolia obovata in relation to crossing type Parameters
Mean ± SE
Germination rate (%) Seedling survival rate (%) Cumulative survival rate (%) Seedling height (cm) Seedling mass (g)
d
Selfed
Outcrossed
36.0±9.6 (5) 20.5±6.8 (5) 0.9±0.4 (5) 7.49±0.47 (34) 3.16±0.42 (34)
58.1±7.3 (5) 32.9±10.1 (5) 10.2±2.4 (5) 9.18±0.21 (158) 4.45±0.19 (158)
0.38±0.14 0.38±0.06 0.91±0.03 0.19±0.07 0.29±0.07
Sample size appears in parentheses.
loci). The 95% confidence interval for the Fis value for all loci ranged between 0.013 and 0.178. The indirect estimates of inbreeding depression for lifetime survival (dw) and late survival (dl) were 0.97 (95% confidence interval ranged from 0.96 to 0.98) and 0.69 (from 0.54 to 0.75), respectively. The secondary selfing rate (r*) estimated from the dw value and primary selfing rate was 0.15. Discussion Inbreeding depression The present study revealed that the highly self-fertilizing M. obovata population exhibited substantial inbreeding depression for early survival. The values of the inbreeding depression for both germination rate and seedling survival rate equaled 0.38. These values may have been underestimated because of mild growing conditions in this study, since less stressful conditions generally mitigate the effects of inbreeding depression (Uyenoyama et al., 1993; Byers Table 2. Results of ANOVA for germination rate, seedling survival rate, and cumulative survival rate of Magnolia obovata in relation to crossing type and family Source Germination rate Crossing type Family Error Seedling survival rate Crossing type Family Error Cumulative survival rate Crossing type Family Error
df
MS
1 4 4
0.1220 0.0561 0.0164
7.43 3.41
0.053 0.131
1 4 4
0.0545 0.0946 0.0036
15.27 26.52
0.017 0.004
1 4 4
0.1340 0.0109 0.0012
113.98 9.23
<0.001 0.027
F
p
182 Table 3. Results of ANOVA for height and mass of 2-year-old Magnolia obovata seedlings in relation to crossing type and family Source Seedling height Crossing type Error Family Crossing type family Error Seedling mass Crossing type Error Family Crossing type family Error
df
MS
1 16.2 4 4 182
14.51 4.90 69.92 4.91 4.89
1 16.2 4 4 182
18.51 2.80 50.67 1.12 4.41
F
p
2.96
0.104
14.24§ 1.00
0.012 0.407
6.62
0.012
45.32§ 0.25
0.001 0.907
The denominator df and MS used to calculate the F ratio for crossing type effect are listed directly under the crossing type-effect line. § The MS for the crossing type family effect was used to calculate the F ratio.
and Waller, 1999). In addition, the previous pollination experiments found that inbreeding depression for embryonic survival was high in the study population (d = 0.76; Ishida et al., 2003). A large M. obovata population on Kyushu island (southern Japan), whose primary selfing rate was 0.82, also exhibited a high level of inbreeding depression for embryonic survival (d = 0.78 ± 0.09; d value ± jackknife standard error using individual maternal trees; K. Ishida, Kansai Research Center, unpublished data). These values of inbreeding depression resulted in strong cumulative inbreeding depression for early survival (de = 0.91), indicating that inbreeding depression for early survival is sufficiently high to maintain traits that facilitate outcrossing in M. obovata (i.e., de > 0.5) according to the theory of mating system evolution based on the cost (inbreeding depression) and the benefit (increased gene transmission) of selfing (Lande and Schemske, 1985). Table 4. Values of heterozygosity and fixation index (Fis) at the late life stage for six allozyme loci in Magnolia obovata Locus
Number of alleles
Observed heterozygosity
Expected heterozygosity
Fis
Adh Aat-1 Aat-2 Dia Lap 6Pgd Average All loci
2 3 2 2 3 2
0.333 0.450 0.217 0.417 0.617 0.250 0.381
0.302 0.478 0.313 0.454 0.622 0.313 0.414
)0.103 0.059 0.307 0.081 0.008 0.201
* The Fis value is marginally significantly greater than zero (p = 0.056).
0.080*
183 The nursery experiment demonstrated significant inbreeding depression for seedling mass (d = 0.29). This inbreeding depression would also potentially increase the magnitude of inbreeding depression for survival or reproduction via half-sib competition, as suggested by Lloyd (1980), since M. obovata is a gap-colonizing species (Kikuzawa, 1987), whose saplings often form a dense clump after gap formation (Mizunaga, 1998). Cumulative inbreeding depression after germination may thus be higher than the cumulative inbreeding depression at the seedling stage, which would be given by dc = 1 ) (1 ) ds)(1 ) dm) = 0.56, where ds and dm are the values of inbreeding depression for seedling survival rate and seedling mass, respectively. Moreover, the allozyme analysis revealed that the indirect estimate of inbreeding depression for late survival was high (dl = 0.69), indicating that inbreeding depression at this late life stage is high in the study population. In a review of inbreeding depression in 54 plant species (Husband and Schemske, 1996), the mean cumulative inbreeding depression for 10 predominantly outcrossing coniferous species (primary selfing rate r < 0.39) was significantly higher than that for 30 predominantly outcrossing angiosperms (no tree species except for one eucalypt). The d values for early survival of M. obovata are as high as those of the 10 conifers: the inbreeding depression for embryonic survival (0.76) was slightly higher than the median (0.74) for the 10 conifers, but the inbreeding depression for germination rate (0.38) was higher than that for all 10 conifers (d = 0.0 ) 0.35). The inbreeding depression for seedling survival (0.38) was also higher than the median (0.21) of inbreeding depression for tree survival (after planting) for the 10 conifers. These examples suggest that, M. obovata exhibits high inbreeding depression for early survival, as is the case for predominantly outcrossing conifers, despite of its high primary selfing rate. These results for inbreeding depression in M. obovata are consistent with the prediction that the long life span and large mature size of a tree increase the consequences of inbreeding depression sufficiently to maintain traits that facilitate outcrossing even under a high self-fertilization rate, a conclusion that is predicted based on previous theories (Lloyd, 1980; Klekowski and Godfrey, 1989; Lande et al., 1994; Morgan, 2001). Some previous studies of other highly self-fertilizing tree species are also available to test this prediction: A large population of Magnolia stellata (Hirayama et al., 2005), which exhibits a high primary selfing rate (r = 0.66) due to geitonogamy, exhibited a high degree of inbreeding depression for embryonic survival (d = 0.70). In a highly self-fertilizing population of Eucalyptus regnans (the Narracan population, r = 0.59; Hardner and Potts, 1997) that self-fertilizes by geitonogamy, the trees also exhibited a high level of inbreeding depression (d = 0.70) for late survival (15 years after planting). Consequently, these high levels of inbreeding depression in highly self-fertilizing species are also consistent with the
184 predictions for inbreeding depression in trees. On the other hand, studies of two predominantly outcrossing coniferous species (Pinus resinosa and Thuja plicata) reported that cumulative inbreeding depression was low (d = 0.06 ) 0.30; Fowler, 1965; Owens et al., 1990; Husband and Schemske, 1996), which is inconsistent with the prediction. Note that T. plicata was formerly classified as a predominantly selfing species, but was considered an outcrossing species by O’Connell et al. (2001). These low levels of inbreeding depression could have been caused by genetic drift resulting from the small population size (Fowler, 1965; Fowler and Morris, 1977; Copes, 1981; El-Kassaby et al., 1994). Nonetheless, the effects of genetic drift might be small in most tree species, since trees generally have large populations due to longdistance gene flow (Chase et al., 1996; Dow and Ashley, 1996; Nason et al., 1998). The level of inbreeding Understanding the genetic mechanisms responsible for the high levels of inbreeding depression in M. obovata requires information on the level of inbreeding, since the level of inbreeding is a key factor in determining the effectiveness of selection at purging deleterious recessive mutations as a result of selective interference among the mutations (Lande et al., 1994). The present study revealed that the indirectly estimated secondary selfing rate was low (r* = 0.15) compared with the primary selfing rate (r = 0.86) in the M. obovata population, even though the estimate would have large statistical variance; given the 95% confidence interval for Fis, the estimated r* ranges from 0.03 to 0.30. This result suggests that strong cumulative inbreeding depression for lifetime survival resulted in a low level of inbreeding at maturity in the M. obovata population despite the high primary selfing rate. In another large M. obovata population in central Japan, the trees exhibited a negative fixation index (high observed heterozygosity compared to the expected heterozygosity at five microsatellite loci) despite a high selfing rate at the seedling stage (rm = 0.71; Isagi et al., 2004); these results provide additional support for the hypothesis of high cumulative inbreeding depression. If a population exhibits a sufficiently high level of inbreeding depression for reproduction (dr) to reduce the opportunity for selfed progenies to reproduce, the proportion of selfed progenies in mature trees that actually contribute to the next generation (rf; denoted the ‘functional’ selfing rate) would be smaller than the secondary selfing rate. The functional selfing rate can be calculated by substituting r* for r and dr for dw in the right-hand side of Equation (1). As inbreeding depression for biomass generally correlates well with that for reproduction (Husband and Schemske, 1996), we can infer the rf value by assuming that the inbreeding depression for seedling mass
185 (dm) equals dr; given r* = 0.15 and dr = 0.29, we obtain rf = 0.11. This example suggests that the functional selfing rate in the M. obovata population is low, and as a result, its mating system can be regarded as predominant outcrossing with respect to the mutation–selection balance for deleterious recessive mutations; a high frequency of deleterious recessive mutations would be maintained in the population compared to the levels in a population with a high functional selfing rate, as if the mutation–selection balance for deleterious recessive mutations is determined under a low primary selfing rate near 0.1 in the population. See figure 1 in Charlesworth and Charlesworth (1987) and figure 1 (cases of the classical model) in Lande et al. (1994) for more details; note that the functional selfing rate in the present study is equivalent to the ‘selfing rate’ in the classical model, as described by Lande et al. (1994). Some highly self-fertilizing eucalypt populations also exhibit a low level of inbreeding, similar to that of M. obovata, suggesting that their high levels of inbreeding depression greatly decrease secondary selfing rates despite the high self-fertilization rates: a large population of a small mallee, Eucalyptus crucis (the Chiddarcooping Hill population), exhibited a high fixation index (Fis = 0.44) at the seedling stage, but a low fixation index (Fis = )0.06) at maturity (Sampson et al., 1988). Note that no biparental inbreeding but a high primary selfing rate can cause a combination of high Fis (nearly 0.5) at the seedling stage and low Fis (nearly zero) maturity. In E. rhodantha, which exhibits an intermediate or high selfing rate at the seed stage due to frequent geitonogamous bird-pollination (rm = 0.41–0.74; Sampson et al., 1989; Potts and Wiltshire, 1997), the trees also exhibited a low fixation index (Fis = )0.12; Sampson et al., 1989), although both their pollination biology and their decreased population size as a result of land clearing may have increased their selfing rates, leading to overestimation of the contribution of inbreeding depression. Genetic basis of inbreeding depression Previous empirical studies have suggested that inbreeding depression for embryonic survival is primarily due to recessive lethal genes in tree species (Husband and Schemske, 1996; Williams and Savolainen, 1996; Wang et al., 1999). If we assume that inbreeding depression for embryonic survival in M. obovata is also caused by typical recessive lethal mutations (a selection coefficient of s = 1.0, and a degree of dominance of h = 0.02; Simmons and Crow, 1977), then the model of inbreeding depression in perennials proposed by Morgan (2001), which incorporates the effects of mitotic mutations via long life span and selective interference among the mutations, can explain the high d value (d =0.76); high levels of inbreeding depression (d > 0.7) can be
186 maintained under a high genomic mutation rate (U = 1.0) even if the primary selfing rate is greater than 0.9 (figure 4 in Morgan, 2001). In contrast, the classical model of inbreeding depression (figure 1 in Charlesworth and Charlesworth 1987) cannot explain such high levels of inbreeding depression even with low levels of inbreeding (where the selfing rate in the model is equivalent to the functional selfing rate in the present study). The substantial inbreeding depression in M. obovata at the seedling stage (cumulative inbreeding depression, dc = 0.56) and inbreeding depression for late survival (dl = 0.67) are consistent with previous hypotheses that inbreeding depression at the late life stage is primarily due to weakly deleterious mutations (Husband and Schemske, 1996; Wang et al., 1999), since such mutations are relatively difficult to purge in the long term as a result of selfing (Charlesworth and Charlesworth, 1987; Barrett and Charlesworth, 1991; Carr and Dudash, 2003). If we assume large contributions of mitotic mutations for late survival, the model of inbreeding depression in perennials predicts that a high level of inbreeding depression (d > 0.5) caused by mildly deleterious mutations (s = 0.1, h = 0.2, U = 1.0) can be maintained even if the primary selfing rate is greater than 0.9 (figure 4 in Morgan, 2001). These examples indicate that previous theoretical models based on the mutation–selection balance for deleterious recessive mutations can explain the high level of inbreeding depression for both the early and the late life stages of M. obovata if we assume the action of both selective interference among deleterious mutations (a low level of inbreeding) and genetic mechanisms that cause a high genomic mutation rate (a high equilibrium frequency of the mutations), such as mitotic mutations. In addition to these genetic mechanisms, synergistic epistasis (Charlesworth, 1998) and overdominance (Charlesworth and Charlesworth, 1987) may also contribute to the high level of inbreeding depression, although there is little evidence for the significance of these mechanisms in the maintenance of inbreeding depression in natural plant populations (Koelewijn, 1998; Carr and Dudash, 2003). Ultimate factors causing high self-fertilization Previous theories have predicted that if the benefit and cost of selfing is determined by its automatic advantage in terms of gene transmission and by the adverse effects of inbreeding depression, then a high level of inbreeding depression would lead to the evolution of traits that facilitate outcrossing (Lande and Schemske, 1985; Uyenoyama et al., 1993). Despite this prediction, M. obovata exhibits both a high primary selfing rate (r = 0.86) and strong inbreeding depression (de = 0.91 and dw = 0.97), implying that the gene transmission advantage operates together with other factors to increase the benefit of selfing, as suggested by Lloyd (1979) and Holsinger (1991). Thus,
187 understanding of the ultimate factors responsible for the high primary selfing rate in M. obovata requires clarification of the adaptive implications (i.e., the costs and benefits) of its pollination system. Magnolia obovata trees can avoid autogamy to some extent as a result of protogyny, but their asynchronous flowering (the coexistence of female-phase and male-phase flowers) causes frequent geitonogamy, leading to a high primary selfing rate (Ishida et al., 2003). On the other hand, Kikuzawa and Mizui (1990) suggest that this asynchronous flowering enables automimicry pollination (Kikuzawa and Mizui, 1990); although the female-phase flowers have neither nectar nor pollen, asynchronous flowering enables female-phase flowers to deceive pollen-feeding insects by displaying mimic (female-phase flowers) and model (male-phase flowers) simultaneously. This automimicry pollination would increase both self- and cross-fertilized ovules by increasing opportunities for interfloral pollination, because the species suffers from a pollen shortage (Kikuzawa and Mizui, 1990; Ishida et al., 2003). Hence, the high self-fertilization rate due to asynchronous flowering can be regarded as a byproduct (cost) of the automimicry pollination system for outcrossing. In addition, mass flowering, which is observed in large M. obovata trees, would also cause the high self-fertilization rate, since mass flowering generally increases geitonogamy (de Jong et al., 1993). As mass flowering increases the attraction of pollinators, increased self-fertilization due to mass flowering can be also regarded as a cost that is incurred to increase outcrossing. Therefore, the cost (inbreeding load) due to self-fertilization may be compensated for by the benefit (increased outcrossing) from the pollination system of M. obovata. If this idea is generally true, the pollination biology of the species is consistent with the theoretical prediction that high levels of inbreeding depression will result in the evolution of traits that facilitate outcrossing. In other tree species, mixed mating due to geitonogamy would also evolve when increased outcrossing compensates for the costs of geitonogamy unless self-incompatibility evolves. Mixed mating due to delayed selfing (autonomous selfing when cross-pollination is insufficient) may also evolve in tree species when this mode of self-pollination provides reproductive assurance (Bernhardt and Thien, 1987; Lloyd, 1992). These modes of mixed mating would help to explain the variation in selfing rate among tree species, even if their levels of inbreeding depression are mostly high. However, floral traits that only facilitate selfing, such as cleistogamy or decreased attraction of pollinators, would remain difficult to evolve in tree species if the predictions about inbreeding depression in trees are generally true. A possible alternative cause of intermediate selfing rates in tree species may be fluctuating inbreeding depression, as suggested by Cheptou and Mathias (2001), although few empirical studies have found fluctuating inbreeding depression in natural plant populations (e.g., Dole and Ritland, 1993).
188 Conclusions The germination test, nursery experiment, and previously published data all demonstrate that highly self-fertilizing M. obovata trees maintain a high level of inbreeding depression during early life stages. The indirect estimate of cumulative inbreeding depression suggests that the trees also maintain a high level of inbreeding depression for later life stages. The inbreeding depression for seedling growth may increase inbreeding depression for late survival by means of half-sib competition. These high levels of inbreeding depression result in a low level of inbreeding at the mature tree stage, thus M. obovata can be regarded as a predominantly outcrossing species with respect to the mutation–selection balance for deleterious recessive mutations. Previous theoretical models of inbreeding depression help to explain the genetic basis of the high level of inbreeding depression that accompanies the high primary selfing rate. These results are consistent with the prediction that the long life span and large mature plant size of trees maintain levels of inbreeding depression sufficiently high to maintain traits that facilitate outcrossing, even under a high primary selfing rate. Understanding the ultimate factors responsible for the predominance of outcrossing in tree species requires further studies of pollination biology with respect to the costs and benefits of selfing in tree populations. Acknowledgements I thank Kazuko Nakamura, Kyoko Tanaka, Yukiko Sakamoto, and Yoshie Ishizuka for their assistance with the germination test, nursery experiment, and allozyme analysis. I also thank Kimiko Hirayama for her invaluable comments on the manuscript. This work was supported by grants from Forestry and Forest Products Research Institute.
References Barrett, S.C.H. and Charlesworth, D. (1991) Effects of a change in the level of inbreeding on the genetic load. Nature 352, 522–524. Barrett, S.C.H. and Eckert, C.G. (1990) Variation and evolution of mating systems in seed plants. In S. Kawano (ed.), Biological Approaches and Evolutionary Trends in Plants. Academic Press, London, pp. 229–254. Barrett, S.C.H. and Harder, L.D. (1996) Ecology and evolution plant mating. Trends Ecol. Evol. 11, 73–39. Barrett, S.C.H., Harder, L.D. and Worley, A.C. (1996) The comparative biology of pollination and mating in flowering plants. Philos. Trans. Roy. Soc. London B 351, 1271–1280. Bernhardt, P. and Thien, L.B. (1987) Self-isolation and insect pollination in the primitive angiosperms: new evaluations of older hypotheses. Plant Syst. Evol. 156, 159–176.
189 Byers, D.L. and Waller, D.M. (1999) Do plant populations purge their genetic load? Effects of population size and mating history on inbreeding depression. Annu. Rev. Ecol. Syst. 30, 479–513. Carr, D.E. and Dudash, M.R. (2003) Recent approaches into the genetic basis of inbreeding depression in plants. Phil. Trans. Roy. Soc. London B 358, 1071–1084. Charlesworth, B. (1998) The effect of synergistic epistasis on the inbreeding load. Genet. Res. Cambridge 71, 85–89. Charlesworth, D. and Charlesworth, B. (1987) Inbreeding depression and its evolutionary consequences. Ann. Rev. Ecol. Syst. 18, 237–268. Charlesworth, D. and Charlesworth, B. (1999) The genetic basis of inbreeding depression. Genet. Res. Cambridge 74, 329–340. Charlesworth, D., Morgan, M.T. and Charlesworth, B. (1990) Inbreeding depression, genetic load, and the evolution of outcrossing rates in a multilocus system with no linkage. Evolution 44, 1469–1489. Chase, M.R., Moller, C., Kesseli, R. and Bawa, K.S. (1996) Distant gene flow in tropical trees. Nature 383, 398–399. Cheptou, P.O. and Mathias, A. (2001) Can varying inbreeding depression select for intermediary selfing rates?. Am. Nat. 157, 361–373. Cheptou, P.O. and Schoen, D.J. (2003) The cost of fluctuating inbreeding depression. Evolution 56, 1059–1062. Copes, D.L. (1981) Isozyme uniformity in western red cedar seedlings from Oregon and Washington. Can. J. For. Res. 11, 451–453. Crawford, D.J. (1983) Phylogenetic and systematic inferences from electrophoretic studies. In S.D. Tanksley and T.J. Orton (eds.), Isozymes in Plant Genetics and Breeding, Part A. Elsevier, Amsterdam, pp. 257–287. Crnokrak, P. and Barrett, S.C. (2002) Purging the genetic load: a review of the experimental evidence. Evolution 56, 2347–2358. Davis, B.J. (1964) Disk electrophoresis II: method and application to human serum proteins. Ann. N. Y. Acad. Sci. 121, 404–427. Dole, J. and Ritland, K. (1993) Inbreeding depression in two Mimulus taxa measured by multigenerational changes in the inbreeding coefficient. Evolution 47, 361–373. Dow, B.D. and Ashley, M.V. (1996) Microsatellite analysis of seed dispersal and parentage of saplings in bur oak, Quercus macrocarpa. Mol. Ecol. 5, 615–627. El-Kassaby, Y.A., Russell, J. and Ritland, K. (1994) Mixed mating in an experimental population of western red cedar, Thuja plicata. J. Hered. 85, 227–231. Fowler, D.P. (1965) Effects of inbreeding in red pine, Pinus resinosa Ait. Silvae Genet. 14, 12–23. Fowler, D.P. and Morris, R.W. (1977) Genetic diversity in red pine: evidence for low genetic heterozygosity. Can. J. For. Res. 7, 343–347. Glemin, S. (2003) How are deleterious mutations purged? Drift versus nonrandom mating. Evolution 57, 2678–2687. Glemin, S., Ronfort, J. and Bataillon, T. (2003) Patterns of inbreeding depression and architecture of the load in subdivided populations. Genetics 165, 2193–2212. Gottlieb, L.D. (1982) Conservation and duplication of isozymes in plants. Science 216, 373–380. Goudet, J. (1995) FSTAT (vers. 1.2): a computer program to calculate F-statistics. J. Hered. 86, 485–486. Hardner, C.M. and Potts, B.M. (1997) Postdispersal selection following mixed mating in Eucalyptus regnans. Evolution 51, 103–111. Hedrick, P.W. (2002) Lethals in finite populations. Evolution 56, 654–657. Hirayama, K., Ishida, K. and Tomaru, N. (2005) Effects of pollen shortage and self-pollination on seed production of an endangered tree, Magnolia stellata. Ann. Bot. 95, 1009–1015. Holsinger, K.E. (1991) Mass-action models of plant mating systems: the evolutionary stability of mixed mating systems. Am. Nat. 138, 606–622. Husband, B.C. and Schemske, D.W. (1996) Evolution of the magnitude and timing of inbreeding depression in plants. Evolution 50, 54–70.
190 Isagi, Y., Kanazashi, T., Suzuki, W., Tanaka, H. and Abe, T. (2000) Microsatellite analysis of the regeneration process of Magnolia obovata Thunb. Heredity 84, 143–151. Isagi, Y., Kanazashi, T., Suzuki, W., Tanaka, H. and Abe, T. (2004) Highly variable pollination patterns in Magnolia obovata revealed by microsatellite paternity analysis. Int. J. Plant Sci. 165, 1047–1053. Ishida, K., Yoshimaru, H. and Ito, H. (2003) Effects of geitonogamy on the seed set of Magnolia obovata Thunb. (Magnoliaceae). Int. J. Plant Sci. 164, 729–735. Johnston, M.O. and Schoen, D.J. (1994) On the measurement of inbreeding depression. Evolution 48, 1735–1741. de Jong, T.J., Waser, N.M. and Klinkhamer, P.G.L. (1993) Geitonogamy: the neglected side of selfing. Trends Ecol. Evol. 8, 321–325. Katsuta, M. (1998) Magnolia Linn. In M. Katsuta, T. Mori and T. Yokoyama (eds.), Seeds of Woody Plants in Japan Angiospermae. Japan Forest Tree Breeding Association, Tokyo, pp. 127–135[In Japanese]. Kennington, W.J. and James, S.H. (1997) The effect of small population size on the mating system of a rare clonal mallee, Eucalyptus argutifolia (Myrtaceae). Heredity 78, 252–260. Kikuzawa, K. (1987) Development and survival of leaves in Magnolia obovata in a deciduous broad-leaved forest in Hokkaido, northern Japan. Can. J. Bot. 65, 412–417. Kikuzawa, K. and Mizui, N. (1990) Flowering and fruiting phenology of Magnolia hypoleuca. Plant Species Biol. 5, 255–261. Klekowski, E.J. (1998) Mutation rates in mangroves and other plants. Genetica 102/103, 325–331. Klekowski, E.J. and Godfrey, P.J. (1989) Ageing and mutation in plants. Nature 340, 389–391. Koelewijn, H.P. (1998) Effects of different levels of inbreeding on progeny fitness in Plantago coronopus. Evolution 52, 692–702. Lande, R. and Schemske, D.W. (1985) The evolution of self-fertilization and inbreeding depression in plants. I. Genetic models. Evolution 39, 24–40. Lande, R., Schemske, D.W. and Schultz, S.T. (1994) High inbreeding depression, selective interference among loci, and the threshold selfing rate for purging recessive lethal mutations. Evolution 48, 965–978. Lloyd, D.G. (1979) Some reproductive factors affecting the selection of self-fertilization in plants. Am. Nat. 113, 67–79. Lloyd, D.G. (1980) Demographic factors and mating patterns in angiosperms. In O.T. Solbrig (eds.), Demography and Evolution in Plant Populations. Blackwell, Oxford, pp. 67–88. Lloyd, D.G. (1992) Self and cross-fertilization in plants. II. The selection of self-fertilization. Int. J. Plant Sci. 153, 370–380. Lynch, M. and Walsh, B. (1998) Genetics and Analysis of Quantitative Traits. Sinauer, Sunderland, MA. Manly, B.F.J. (1997) Randomization, Bootstrap and Monte Carlo Methods in Biology. 2nd edn. Chapman & Hall, London. Mizunaga, H. (1998) Effects of disturbances on large patch formation of Magnolia obovata, a deciduous broad-leaved tree – field investigation and simulation. J. Sustainable For. 6, 251–266. Morgan, M.T. (2001) Consequences of life history for inbreeding depression and mating system evolution in plants. Proc. Roy. Soc. London B 268, 1817–1824. Nason, J.D., Herre, E.A. and Hamrick, J.L. (1998) The breeding structure of a tropical keystone plant resource. Nature 391, 685–687. Nei, M. (1987) Molecular Evolutionary Genetics. Columbia University Press, New York. O’Connell, L.M., Viard, F., Russell, J. and Ritland, K. (2001) The mating system in natural populations of western redcedar (Thuja plicata). Can. J. Bot. 79, 753–756. Orstein, I. (1964) Disk electrophoresis. I: Background and theory. Ann. N.Y. Acad. Sci. 121, 321–349. Owens, J.N., Colangeli, A.M. and Morris, S.J. (1990) The effect of self-, cross-, and no pollination on ovule, embryo, seed, and cone development in western red cedar (Thuja plicata). Can. J. For. Res. 20, 66–75. Potts, B.M. and Wiltshire, R.J.E. (1997) Eucalypt genetics and genecology. In J. Williams and J. Woinarski (eds.), Eucalypt Ecology. Cambridge University Press, Cambridge, pp. 56–91.
191 Ritland, K. (1990) Inferences about inbreeding depression based on changes of the inbreeding coefficient. Evolution 44, 1230–1241. Roff, D.A. (2002) Inbreeding depression: tests of the overdominance and partial dominance hypothesis. Evolution 56, 768–775. Sampson, J.F., Hopper, S.D. and James, S.H. (1988) Genetic diversity and the conservation of Eucalyptus crucis Maiden. Aust. J. Bot. 36, 447–460. Sampson, J.F., Hopper, S.D. and James, S.H. (1989) The mating system and population genetic structure in a bird-pollinated mallee, Eucalyptus rhodantha. Heredity 63, 383–393. Sano, J. (1997) Age and size distribution in a long-term forest dynamics. For. Ecol. Manage. 92, 39–44. Satoo, S. (1992) Propagation of 30 broad-leaved trees species in Hokkaido by seedling – practical results from the nursery of the Tokyo University Forest in Hokkaido. Bull. Tokyo Univ. For. 87, 89–128. [In Japanese, with English abstract]. Shiraishi, S. (1988) Inheritance of isozyme variations in Japanese Black Pine, Pinus thunbergii Parl. Silvae. Genet. 37, 93–100. Simmons, M.J. and Crow, J.F. (1977) Mutations affecting fitness in Drosophila populations. Annu. Rev. Genet. 11, 49–78. Ueda, K. (1986) A nomenclatural revision of the Japanese Magnolia species (Magnoliaceae) together with two long-cultivated Chinese species. I. M. hypoleuca. Taxon 35, 340–344. Uyenoyama, M.K., Holsinger, K.E. and Waller, D.M. (1993) Ecological and genetic factors directing the evolution of self-fertilization. Oxford Surv. Evol. Biol. 9, 327–381. Wang, T., Hagqvist, R. and Tigerstedt, M.A. (1999) Inbreeding depression in three generations of selfed families of silver birch (Betula pendula). Can. J. For. Res. 29, 662–668. Williams, C.G. and Savolainen, O. (1996) Inbreeding depression in conifers: implications for breeding strategy. For. Sci. 42, 102–117. Wright, S. (1969) The Theory of Gene Frequencies, Evolution and the Genetics of Populations. 2University of Chicago Press, Chicago. Yasukawa, S., Kato, H., Yamaoka, R., Tanaka, H., Arai, H. and Kawano, S. (1992) Reproductive and pollination biology of Magnolia and its allied genera (Magnoliaceae). I. Floral volatiles of several Magnolia and Michelia species and their roles in attracting insects. Plant Species Biol. 7, 121–140.
