The link between selfing and greater dispersibility in a heterocarpic Asteraceae

1 Running head: Breeding systems and dispersibility 1 2 The link between selfing and greater dispersibility in a heterocarpic Asteraceae1 3 4 R. Berjano2,3,4, N. L. Rodríguez-Castañeda3, P. L. Ortiz3, M. A. Ortiz3 & M. Arista3 5 2Departamento de Botánica, Biología Vegetal y Fisiología Vegetal. Universidad de 6 Córdoba. 14071. Córdoba, Spain. 7 8 3Departamento de Biología Vegetal y Ecología. Universidad de Sevilla. 9 Apdo. 1095, 41080 Sevilla, Spain. 10 11 12 13 14 15 16 17 18 19 20 21 22 Key words: breeding systems; coastal herb; dispersal evolution; fruit dimorphism; hand 23 pollination; heterocarpy; Hypochaeris; imbreeding depression; plume loading; 24 Mediterranean. 25 26 27 28 29 30 31 32 1Manuscript received _______; revision accepted _______ 33 34 4Author for correspondence ([email protected]) 35 36 2 SUMMARY 37 PREMISE OF THE STUDY: Although evolutionary link between breeding systems 38 and dispersibility has been proposed, to date empirical data and theoretical models on 39 plants show contrasting trends, sometimes associating selfing and non-dispersal, and 40 others selfing and dispersal. 41 METHODS: We tested two competing hypotheses for the association between 42 breeding systems and dispersibility in the heterocarpic Hypochaeris salzmanniana 43 (Asteraceae) by using both an experimental approach and surveys in five natural 44 populations occurring in a gradient of pollinator availability over two years. 45 KEY RESULTS: H. salzmanniana produced two types of fruits, beaked (BF) and non- 46 beaked (NBF), differing in dispersal ability. BF were lighter and showed a lower 47 dropping velocity and higher dispersal distance than NBF. Potential for long dispersal, 48 measured as BF ratio per head, had high narrow-sense heritability. Greater dispersibility 49 and selfing ability were linked at all the scales studied. Both selfed BF and NBF fruits 50 exhibited longer plumes and lower plume loading than outcrossed, characteristics 51 promoting farther dispersal. Natural populations with higher percentage of self- 52 compatible plants showed higher BF ratio. Moreover, selfing led to higher BF ratio than 53 outcrossing. 54 CONCLUSIONS: 55 The avoidance of inbreeding depression seems to be the most plausible selective 56 pressure for greater dispersibility traits of selfed seeds. Furthermore, the ability to 57 modulate the BF ratio and thus, the potential for long dispersal of offspring, based on its 58 selfed or outcrossed origin could be advantageous, and therefore selected, under 59 unpredictable pollination environments that favors higher dispersive selfers that 60 overcome both pollen limitation and inbreeding avoidance. 61 3 INTRODUCTION 62 Breeding systems and seed dispersal are key factors modeling plant metapopulation 63 dynamics and species’ distributions, and can be affected by several factors that are 64 omnipresent in the current global change, as habitat loss, fragmentation or pollinator 65 decline. Breeding system and dispersal ability are anatomically and ontogenetically 66 linked, and selective pressures on breeding systems can generate consequences for 67 dispersal, and vice versa (Primack, 1987; Rubio de Casas et al., 2012; Auld and Rubio 68 de Casas, 2013). Both reproductive traits impact gene flow, genetic diversity and 69 colonization ability of flowering plants, and thus may affect potentiality for adaptation 70 of their populations; therefore, an evolutionary association between both reproductive 71 traits could be expected (de Waal et al., 2014). However, the association between 72 breeding systems and dispersal has not a clear pattern, and opposing links (low or high 73 dispersal and selfing) are predicted (Auld & Rubio de Casas, 2013; de Waal et al, 74 2014). 75 Inbreeding avoidance and environmental stochasticity are two explanatory hypotheses 76 for the correlated evolution of these traits (Olivieri et al., 1995; Perrin and Goudet, 77 2001; Guillaume and Perrin, 2006; Ronce, 2007; Auld and Rubio de Casas, 2013). In 78 species suffering high inbreeding depression (i.e. fitness reduction due to the expression 79 of homozygote recessive alleles; Charlesworth & Willis, 2009), dispersal would reduce 80 the chance of mating between genetically close individuals (Waser, 1993), which could 81 constitute a selective force favoring high dispersibility of selfed progeny. Several 82 theoretical studies predict that dispersal is favored by high inbreeding depression 83 (Perrin & Mazalov, 1999; Perrin & Goudet, 2001; Roze & Rousset, 2005). 84 In contrast, theoretical models based on pollinator availability, predicted a strict 85 association between obligate outcrossing and greater dispersibility under random 86 4 pollination fluctuations (Cheptou and Massol, 2009). Stochasticity in pollination 87 regimes, which are widespread in natural populations (Burd 1994) is central to 88 interpreting the association of mating system and dispersal traits (Massol & Cheptou, 89 2011a). In their theoretical models, Cheptou and Massol found that dispersal was never 90 associated with selfing (Cheptou and Massol 2009; Massol & Cheptou, 2011a). These 91 authors argued that uncertainty in pollination triggers a lethal cost for non-dispersing 92 outcrossers, which could be overcome either by dispersing, or by selfing (Massol & 93 Cheptou, 2011b). Both mechanisms have costs: the former implies paying the cost of 94 dispersal, and the latter, paying the cost of inbreeding. For that reason, the authors argue 95 that selfers with greater dispersibility could not evolve since they would be 96 outcompeted by mutants that disperse or self-fertilize less (Massol & Cheptou, 2011b). 97 Heterocarpic species constitute excellent systems to test dispersibility and mating 98 systems. Heterocarpic species produce at least two fruit types differing in dispersal 99 ability, which have been seen as an adaptive strategy for plants dealing with 100 heterogeneous environments (Lloyd, 1968). Heterocarpy, described in 18 families, is 101 especially common in Asteraceae family, which accounts over the half of the described 102 species (Imbert, 2002). In heterocarpic Asteraceae, the link between selfing- 103 dispersibility may be due to different trait associations. Within the flower head, less 104 dispersible fruits are always located at the periphery and, according to the centripetal 105 phenological flowering pattern, they are produced first (Imbert, 2002). These less 106 dispersive fruits are only produced at the first flower rows within the flower head and 107 are much less numerous than inner fruits. In fact, the number of outer flowers, 108 controlled by the number of parastichies (i.e. the number of phyllotactic spirals of 109 flowers that composes the flower head), is expected to show no genetic variance (Imbert 110 & Ronce, 2001). In radiate Asteraceae species, the two types of fruits are associated 111 5 with different flower types: outer female ray flowers and inner hermaphrodite disk 112 flowers (Venable and Levin, 1985). Thus, outer fruits come from female flowers, which 113 necessarily have higher outcrossing rates than inner ones (Marshall and Abbott, 1982, 114 1984), and this leads to the association between outcrossing and non-dispersibility (but 115 see Gibson, 2001; Gibson, Tomlinson, 2002 for similar outcrossing rates in ray and disk 116 flowers). In contrast, in other heterocarpic Asteraceae with only hermaphrodite flowers, 117 contrasting associations have been found in relation to mating system and flower 118 position. Thus, in Carduus species, protandry along with centripetal head phenology, 119 makes selfing more likely in outer flowers, leading to the link between selfing and non- 120 dispersibility (Olivieri et al., 1983; Ravigné et al., 2006). However, in Crepis sancta 121 (L.) Babc., with similar protandy and head phenology, higher outcrossing levels using 122 allozymes markers have been found in outer flowers, giving rise to the link between 123 outcrossing and non-dispersibility (Cheptou et al., 2001). The association between 124 selfing and greater dispersibility is also found in a multi-species study of Asteraceae 125 from South Africa (de Waal et al., 2014). 126 Evidence for the link between selfing and non-dispersibility comes mainly from 127 cleistogamous or amphicarpic plants exhibiting mixed-mating systems (Cheplick, 128 1987). In these species, diaspores from cleistogamous or basal flowers have no dispersal 129 mechanism and can even be buried just beneath the mother plants, whereas those from 130 chasmogamous or aerial flowers possess some dispersive mechanism. This happens in 131 several grasses (Cheplick, 1993a, b; Clay, 1983; McNamara and Quinn, 1977) and also 132 in cleistogamous Asteraceae, as Centaurea melitensis L., in which less dispersible fruits 133 are associated with cleistogamous heads, which has been considered as an adaptation to 134 unstable environment (Porras and Muñoz, 2000). Selfing and reduced dispersal 135 6 potential as a consequence of an increase in fruit size was also found in the non- 136 cleistogamous Hypochaeris radicata L., (Mix et al., 2006). 137 No doubt further empirical work is needed to get a deeper insight into the links between 138 breeding system and dispersibility, a need also claimed by other authors (Auld and 139 Rubio de Casas, 2013; Iritani and Cheptou, 2017). Very few studies have tested for 140 differences in seed dispersal potential between closely related selfing vs outcrossing 141 populations (but see Darling et al., 2008), so no empirical generalization is currently 142 possible. 143 Here, we documented the association between selfing and dispersibility by monitoring, 144 for two consecutive years, five natural populations of Hypochaeris salzmanniana DC. 145 (Asteraceae) distributed along a gradient of pollinator availability, as well as by using 146 an experimental approach. It has a sporophytic self-incompatibility system (Ortiz et al., 147 2006), by which self-fertilization is prevented by the inherited capacity of flowers to 148 reject incompatible pollen (including its own), based on the sporophitic (i.e diploid) 149 genetic control of pollen and pistil by means of recognition and associated self- 150 rejection processes (de Nettancourt, 2001). 151 Populations of H. salzmanniana are self-incompatible in North Africa and presumably 152 lost its self-incompatibility system when the species migrated to SW Spain, where they 153 show a mixture of self-incompatible, partially self-compatible and fully self-compatible 154 plants (Ortiz et al., 2006). Self-compatible plants are able to self-fertilize automatically 155 even without any pollinator attendance (Arista et al. 2017). In SW Spain, pollination 156 environment varies at a few kilometers distance from west toward east due to extreme 157 winds in the Strait of Gibraltar area, making spatial differences in pollinator 158 environment higher than temporal ones (Arista et al., 2017). Populations occupying 159 areas with extreme winds (pollination-limited environments) are mainly composed of 160 7 self-compatible individuals and show low genetic diversity (Ortiz et al., 2006; Arista et 161 al., 2017). This geographic variation in self-incompatibility has been associated with 162 reproductive assurance due to lack of mate availability caused by both low genetic 163 diversity and unfavorable pollinator environment (Ortiz et al., 2006; Arista et al., 2017). 164 Hypochaeris salzmanniana is an endangered species with a distribution area restricted 165 to coastal sand dunes, habitat that have suffered in Spain an impressive reduction in the 166 last 40 years mainly due to real estate constructions along the coastline (Ortiz et al., 167 2003). We chose H. salzmanniana as a model system to test the association between 168 selfing ability and dispersibility because it occurs along a pollination-environment 169 gradient, and exhibits variability in its reproductive system that shows clear patterns of 170 inheritance (progeny after selfing are mostly self-compatible; Arista et al., 2017). 171 Moreover, the species shows inbreeding depression, which varies spatially and 172 temporally in intensity (Arista et al., 2017). Lastly, plants produce two types of fruits 173 that presumably differ in dispersal distance and fruit-morph ratio can give an estimator 174 of dispersibility of the progeny. Changes in fruit-morph ratio may affect fitness through 175 changes in the dispersal rate of the progeny (Venable, 1985; Olivieri et al., 1995), and it 176 could evolve if it is heritable. The combination of mixed mating systems and 177 heterocarpy could give rise to either of two contrasting associations between selfing and 178 lower/greater dispersibility, as mentioned above. On one hand, given the high level of 179 inbreeding depression of H. salzmanniana, a strong selection for greater dispersibility 180 of selfed progeny is expected to lower the risk of inbreeding. On the other hand, under 181 conditions of pollination environment fluctuations, outcrossers would benefit from 182 dispersal, and consequently the link between selfing and non-dispersal (or lower 183 dispersal) is to be expected. 184 We use a combination of observations of natural populations and experimental hand 185 8 pollinations to test for heritability and changes in dispersibility in relation to breeding 186 system (i.e. among-individual level) and pollen source (i.e. within-individual level). 187 Specifically, we want to answer the following questions: 1) Is fruit-morph ratio a 188 heritable trait? This is a necessarily condition for a trait to evolve 2) Does selfing affect 189 fruit production and fruit traits affecting dispersal ability? If so, is this a consequence of 190 inbreeding depression or a strategy to avoid it? 3) Are there differences in fruit-morph 191 ratios between self-compatible and self-incompatible plants? and 4) Does pollen source 192 (self vs. cross) affect fruit-morph ratio in self-compatible plants? The ability to 193 increased fruit-morph ratio after selfing could represent an advantage to avoid 194 inbreeding depression, while an increased fruit morph ratio after outcrossing could be 195 seen as a selective pressure favoring outcrossers in unfavorable pollination 196 environments. 197 MATERIAL AND METHODS 198 Study species and populations studied—Hypochaeris salzmanniana is an annual 199 species belonging to the tribe Lactuceae from the Asteraceae family, endemic to a 200 restricted area in both sides of the strait of Gibraltar (SW Spain – NW Morocco). Its 201 fruit heads produce two wind-dispersed fruit types (achenes with plumes; Fig. 1): non- 202 beaked plumed fruits at the periphery of the head, and beaked plumed fruits at the 203 center. 204 Field study was carried out for two consecutive years in five natural populations of H. 205 salzmanniana occurring along an environmental cline in southwestern Spain (Cádiz 206 province). These populations were from West to East: Conil (36.21º N, 6.06º W), Caños 207 (36.20º N, 6.05º W), Barbate, (36.21º N, 5.93º W), Zahara (36.14º N, 5.86º W), and 208 Tarifa (36.13º N, 5.84º W). All of them occur on the first sand dune of the coast, with 209 9 the exception of Barbate, where plants grow in a fixed dune under a forest of Pinus 210 pinea L., about 2 km from the sea shore. These populations show important differences 211 in self-incompatibility, with self-incompatible plants decreasing in frequency towards 212 the east due to an increase of winds during flowering which affect pollinator activity 213 (Arista et al., 2017). Self-incompatibility of these populations was studied in 2002, 214 2003, 2014 and 2015; in all years, Tarifa and Zahara were composed mainly of self- 215 compatible plants, while Conil, Caños and Barbate of self-incompatible ones (Ortiz et 216 al., 2006; Arista et al., 2017). 217 Fruit traits affecting dispersal ability and differences between the two types of 218 fruits— Traits that could affect fruit dispersal ability were recorded in the two types of 219 fruits by measuring the lengths of the plume, the beak and the fruit, the opening 220 diameter of the plume, the width of the fruit and the total weight of the fruit (Fig. 1). 221 These traits were measured in 6-15 fruits coming from 18 plants from Zahara and 222 Barbate populations (in total, 75 beaked and 77 non-beaked, hereafter BF and NBF). 223 Plume loading was calculated as fruit weight/plume area (Andersen, 1993). Plume 224 shape was assumed as the lateral surface of an inverted cone, with the cone base area as 225 the plume area. Note that plume loading is not calculated with plume length. 226 To determine the relative wind dispersal ability of both types of fruits, we used two 227 estimates: dropping velocity and dispersal distance. Dropping velocity was calculated 228 by taking the fall time of each fruit at 2m height. The same observer recorded the time 229 in each case. Although fall time is considered a proxy for dispersal distance in many 230 studies (e.g. Cody and Overton, 1996; Fresnillo and Ehlers, 2008), we also estimated 231 differences in dispersal distance between fruits by using a wind tunnel. The wind tunnel 232 consisted of a transparent tube 8m long, with an industrial fan placed at one end, 233 producing a simulated wind source of 2.5 m/s measured by a Brunton Sherpa 234 16 Beaked fruits were produced at the center of the head, and their number was also 383 extremely variable (range: 1-96; mean: 37). The number of both types of fruits were 384 markedly correlated with total fruit production (P < 0.001), but the correlation 385 coefficient of BF was higher (ρ = 0.81; n = 74) than that of NBF (ρ = 0.63; n=74). 386 The mean number of BF per head of the progeny (F2) were directly correlated with 387 those of their F1 mother plants (ρ=0.54, n=74, P < 0.001). Similarly, a significant 388 correlation was found between the mean number of NBF per head of F1 and F2 (ρ=0.61, 389 n=74, P < 0.001). Mean BF ratio of F1 plants was 0.579 ± 0.022, while for F2, mean BF 390 ratio was 0.662 ± 0.016. The BF ratio of F2 plants were positive and significantly 391 correlated with those of their F1 mothers (ρ=0.62, P < 0.001, n=74). A significant linear 392 regression BF ratio of F2 to F1 showed that heritability (h2) was 0.51 (Fig. 3). 393 Dispersibility and its relationship with breeding systems 394 In self-compatible plants, the type of pollen received by a flower head (self or outcross) 395 influenced BF ratio of the fruit head, as well as fruit traits affecting dispersibility (fruit 396 weight and plume length). Fruit production was slightly higher in outcrossed heads than 397 in selfed ones, but this result was only marginally significant (F1, 118= 3.55, P = 0.062). 398 Although the production of each fruit type was similar between selfing and outcrossing 399 treatments (NBF: F1, 118= 3.06, P = 0.083; BF: F1, 118= 2.62, P = 0.110), fruit heads 400 produced by selfing showed significantly higher BF ratio than those produced by 401 outcrossing (Z= 2.46, P = 0.014; Fig. 4A). 402 In relation to fruit traits, after accounting for genetic family, fruits produced after selfing 403 were markedly lighter than after outcrossing (F1, 540= 117.03, P < 0.001), and BF were 404 always lighter than NBF (F1, 540= 275.14, P < 0.001); the fruit type by cross type 405 interaction was not significant (P > 0.05). BF had a mean weight of 1.4 ± 0.029 mg 406 after selfing, and of 1.8 ± 0.027 mg after outcrossing, while NBF had 2.04 ± 0.35 mg 407 17 and 2.46 ± 0.26 mg after selfing and outcrossing respectively. Plume length of fruits 408 after selfing were also significantly different than after outcrossing, but the tendency 409 was opposite to that found in fruit weight: after accounting for genetic family, BF had 410 always longer plume lengths than NBF (F1, 114= 35.67, P < 0.001), and plume length 411 was significantly longer in fruits after selfing (13.44 ± 0.18 mm) than after outcrossing 412 (12.49 ± 0.17 mm; F1, 114= 7.28, P = 0.008), a trend which was maintained in both BF 413 and NBF types (fruit type by cross type interaction not significant; P > 0.05). 414 Models for total fruit production, number of BF and number of NBF as response 415 variables revealed that self-incompatibility was not a significant predictor (P > 0.05). 416 Average BF ratio was higher in self-compatible plants, but the GLMM model did not 417 show significant differences among incompatibility types (P > 0.05; Fig. 4B). 418 DISCUSSION 419 This work provides empirical evidence about the differences in dispersal ability of fruits 420 in the annual heterocarpic Hypochaeris salzmanniana, and about the link between 421 selfing and greater dispersibility at different levels. In this species, both types of fruits 422 differed in beak presence, and beaked fruits (BF) were lighter, and produced longer 423 plumes with a wider aperture diameter than NBF. Plume loading affected dropping 424 velocity, as has been previously found in other Asteraceae (Andersen, 1993), and also 425 dispersal distance of fruits. While analyzing traits separately, fruit weight, plume 426 diameter and plume length affected dropping velocity. In contrast, the diameter of the 427 plume was the sole trait affecting fruit dispersal distance. Fruit weight was negatively 428 correlated with both plume length and plume diameter, leading to the following 429 tendencies: the lighter the fruits, the larger the length and diameter of plume, and thus 430 the lower the dropping velocity and the longer the dispersal distance. In fact, we have 431 18 found experimentally that BF had a lower dropping velocity and a longer dispersal 432 distance than NBF. This suggests that in the field, BF are responsible for the long- 433 distance dispersal, while NBF disperse more locally. Thus, the percent of beaked fruits 434 (BF ratio) is a good estimate of the potential capacity for long dispersal of a plant. Our 435 study also demonstrates that fruit dropping velocity is negatively correlated to dispersal 436 distance, and thus dropping velocity is a useful trait as a surrogate of dispersal distance 437 in wind-dispersed species (Cody and Overton, 1996; Fresnillo and Ehlers, 2008; de 438 Waal et al., 2014). 439 BF ratio in H. salzmanniana is a trait with a high narrow-sense heritability, as h2 for 440 mid-parent regression was 0.51. From this result, we can conclude that phenotypic 441 variance for dispersal ability has a genetic component. However, we must take into 442 account that genetic variance depends on allelic frequencies, and thus, estimation of h² 443 is population-dependent. As our experimental design included crosses of plants from 444 two different populations, the heritability level found in H. salzmanniana could be 445 overestimated. In the other sole species to our knowledge in which the heritability of a 446 dispersal ratio has been measured, Crepis sancta, narrow-sense heritability was higher 447 than 0.2, despite the ontogenetic contingency observed in this species (Imbert, 2001). 448 This suggests that dispersibility is heritable in Asteraceae. 449 The most important result we have found is a strong link between selfing and greater 450 dispersibility at two scales. Selfing leads to 1) a higher proportion of long-dispersible 451 progeny (i.e. higher BF ratio), and 2) fruits with traits that increase dispersal ability. In 452 our study species, selfed and outcrossed fruits differed markedly in mass, selfed fruits 453 being much lighter but with longer plumes, irrespective of position in the head. Given 454 that fruit weight and plume diameter were negatively correlated, and that plume loading 455 was the most important variable affecting dispersibility, lighter fruits dispersed at longer 456 19 distances. Thus, both selfed BF and NBF dispersed at longer distances than those from 457 outcrossing. Differences in mass could be due to inbreeding depression, as this species 458 shows high levels of inbreeding depression at different life-stages as well as throughout 459 the total life cycle (Arista et al., 2017); but inbreeding depression cannot account for the 460 longer plumes of selfed fruits as they would be more costly. Iritani and Cheptou (2017) 461 proposed that, alternatively to the inbreeding depression interpretation, the lower size of 462 selfed seeds could be an adaptive trait mediating differential seed dispersal. In our 463 studied species, the fact that plume length of selfed fruits was significantly longer than 464 that of outcrossed fruits, promoting a higher dispersibility, also supports the adaptive 465 significance of the greater dispersal ability of selfers. 466 Although the numbers of NBF and BF fruits were similar after both pollination 467 treatments in SC plants, we found that the BF ratio of H. salzmanniana was 468 significantly higher in selfed than in outcrossed heads. This suggests that, after selfing, 469 plants could reallocate resources to increase their BF ratio. Changes in BF ratio due to 470 resource reallocation has been documented in other Asteraceae, leading to an increased 471 dispersal rate of their progeny under stressful conditions (Imbert & Ronce, 2001) or to a 472 decreased wind dispersal rate with increased density (Baker and O’Dowd, 1982; Ruiz de 473 Clavijo and Jimenez, 1998). However, in both cases it is not possible to reject 474 developmental constraints as a source of variation in fruit morph proportions (Imbert & 475 Ronce, 2001). In fact, in the tribe Lactuceae, the number of peripheral flowers in a head 476 is expected to be bounded to 13 (Imbert & Ronce, 2001). However, in H. salzmanniana, 477 we found a wide range of NBF (up to 53), and significant correlations between 478 peripheral NBF and total fruits, and between NBF of mother plants and progeny were 479 observed, suggesting maternal effects. Thus, in our study, developmental constraints do 480 not seem the main source of variation in fruit morph proportions. Rather, the differences 481 20 in BF ratio could be seen as an ability to modulate the potential for long dispersal of 482 offspring based on its selfed or outcrossed origin. This ability would be advantageous, 483 and therefore could be selected, under unpredictable pollination conditions. However, 484 further observations in fruit-development phenology within the flower heads would be 485 necessary to check whether a resource allocation adjustment among fruit types is acting. 486 In any case, the adaptive significance of the response of BF ratio to pollen source is 487 difficult to interpret, since fruit types could differ by other characteristics, as dormancy 488 (Venable and Levin 1985; Picó et al., 2003), that could provide distinct ecological 489 differences to each fruit type (Gibson, 2001). 490 The results recorded in five natural populations of H. salzmanniana over two years also 491 uphold the link between selfing and dispersibility found glasshouse experiments. 492 Populations with a high proportion of SC plants and worse pollinator environment due 493 strong winds (Zahara and Tarifa) showed higher BF ratio than more self-incompatible 494 populations (Conil and Barbate; Ortiz et al., 2006; Arista et al., 2017). Thus, high 495 potential for long dispersal appears to be associated with self-compatibility and worse 496 pollinator environment. These results do not support the local adaptation hypothesis 497 which predicts that selfed progeny will perform better in the local area because they 498 conserve a genetic combination adapted to the present environment (Schmitt and 499 Gamble, 1990). Our results are also opposed to those of the extension of Cheptou- 500 Massol models (Cheptou and Massol, 2009; Massol and Cheptou, 2011; Sun and 501 Cheptou, 2012). In these models, self-fertilization evolves when environment limits 502 pollination, as has been suggested in natural populations of H. salzmanniana (Arista et 503 al., 2017). However, those theoretical models assume a decrease in inbreeding 504 depression in self-pollinating plants, and thus a low pressure for dispersal. This situation 505 is not found in H. salzmanniana where self-compatible plants suffer inbreeding 506 21 depression (Arista et al., 2017). Given the high cost of inbreeding in this species, the 507 greater dispersibility of selfed progeny could represent a strategy to avoid that cost. 508 Moreover, seedlings resulting from inbreeding are closely related; therefore, the 509 avoidance of sib-competition cannot be ruled out as a selective pressure for dispersal 510 (Cheplick, 1993a). In fact, the consequences of inbreeding depression for the evolution 511 of dispersal cannot be understood without taking into account its complex interactions 512 with sib competition (Ronce, 2007). 513 The link of selfing and greater dispersibility found in H. salzmanniana is also supported 514 by a theoretical study which predicts that selfing selects for and is selected for an 515 increased seed dispersal, although the evolutionary outcome is strongly influenced by 516 the relative cost of pollen vs. seed dispersal (Ravign et al., 2006). More recently, Iritani 517 and Cheptou (2017) developed several theoretical models, with increasing complexity, 518 for the evolution of mating system and differential seed dispersal in metapopulations, 519 incorporating heterogeneous pollination, dispersal cost, outcrossing cost and 520 environment-dependent inbreeding depression. In those models dealing with the joint 521 evolution of multiple traits, evolutionary patterns not predicted on previous simpler 522 models (Cheptou and Massol, 2009; Massol and Cheptou, 2011a) arise and show that, 523 when selfing and dispersal evolve together, evolution would lead to higher or equal 524 dispersal rate for selfed seeds compared to that for outcrossed (Iritani and Cheptou, 525 2017). 526 Within-species variability in dispersal traits indicates that dispersal strategies can be 527 evolutionarily labile (Van Den Elzen et al., 2016). In our study species, the heritability 528 of dispersibility and the existence of variability in BF ratio in natural populations 529 suggest that changes in long-distance dispersibility over a few generations of selection 530 are possible in this species, as demonstrated in Crepis sancta (Cheptou et al., 2008). 531 22 The heritable character of self-compatibility in H. salzmanniana (Arista et al., 2017) 532 also implies that more dispersible fruits produced after selfing give rise preferentially to 533 self-compatible plants, which would be more successful in founding new colonies than 534 self-incompatible ones, due to reproductive assurance by uniparental reproduction 535 (Pannel, 2015). This is crucial for an annual plant whose populations shows high 536 stochasticity in pollinator conditions. 537 In conclusion, our study shows that the two types of fruits of H. salzmanniana differ in 538 dispersal availability and that BF ratio is a heritable trait related with long dispersal 539 potential. The link found experimentally between selfing and greater BF ratio has also 540 been recorded in natural populations over two years, and seems to have been selected as 541 a way of avoiding inbreeding depression of selfers. Self-fertilization in H. salzmanniana 542 could enhance the colonization of vacant habitat patches, sustaining metapopulations 543 (Olivieri et al., 1983, 1995; Pannell and Barrett, 1998, Pannell et al., 2015), and the 544 potential expansion of the species’ geographic distribution at range margins (Thomas et 545 al., 2001; Travis and Dytham, 2002; Darling et al., 2008). The enhanced dispersal rates 546 observed at range margins, as detected in Abronia umbellata (Nyctaginaceae; Darling et 547 al., 2008) and also in exotic ranges of invasive species such as de congeneric 548 Hypochaeris glabra (Martín-Fores et al., 2018) confirms the importance of dispersal 549 modeling plant distributions. Moreover, dispersal plays a key role in species’ responses 550 disturbances (Harveargres and Eckert, 2014), as habitat loss or fragmentation, which 551 represent current conservation problems that involve changes in species’ distributions 552 (Sexton et al., 2009). This work contributes to understanding the link between mating 553 systems and dispersal adaptation in an endangered species inhabiting a fragile habitat 554 with unfavorable pollinator environment, and highlight the association between selfing 555 and higher dispersibility. Our contribution can also give insights to understand the joint 556 23 evolution of selfing and dispersibility and its influence plant distributions in ways that 557 are important in the current context of environmental challenges, such as the decline of 558 pollinators and habitat loss. 559 560 ACKNOWLEDGEMENTS 561 The authors thank P. O. Cheptou, two anonymous reviewers and the associate editor S. 562 Kephart for critical comments that substantially improved the Ms. We also thank “SGI 563 Herbario" and "SGI Invernadero" of the University of Sevilla and the DNA Bank Herb 564 SEV for lab and glasshouse facilities, and the Natural Park “La Breña y Marismas de 565 Barbate” for permitting us perform this study. This work was supported with FEDER 566 funds and grants from the Spanish MINECO (CGL2012-33270; CGL2015-63827; 567 CGL2015-66161-P), and from the MUBA Master’s program of Seville University 568 569 REFERENCES 570 571 ANDERSEN, M.C. 1993. Diaspore morphology and seed dispersal in several wind- 572 dispersed Asteraceae. American Journal of Botany 80: 487–492. 573 ARISTA, M., R. BERJANO, J. VIRUEL, M.A. ORTIZ, M. TALAVERA, and P.L. ORTIZ. 2017. 574 Uncertain pollination environment promotes the evolution of a stable mixed 575 reproductive system in the self-incompatible Hypochaeris salzmanniana 576 (Asteraceae). Annals of Botany 120:447-456 577 AULD, J.R., and R. RUBIO DE CASAS. 2013. The correlated evolution of dispersal and 578 mating-system traits. Evolutionary Biology 40: 185–193. 579 BATES, D., MAECHLER, M., BOLKER, B., and WALKER, S. 2015. Fitting Linear Mixed- 580 Effects Models Using lme4. Journal of Statistical Software 67: 1-48. 581 BAKER, G. A. and D.J. O’DOWD, 1982. Effects of parent plant density on the production 582 of achene type in the annual Hypochœris glabra. Journal of Ecology 70: 201–215. 583 BURD, M. 1994. Bateman principle and plant reproduction: the role of pollen limitation 584 in fruit and seed set. Botanical review 60: 83-139. 585 CHARLESWORTH D. and WILLIS J.H. 2009. The genetics of inbreeding depression, 586 Nature Reviews Genetic: 10: 783-796. 587 24 CHEPLICK, G.P. 1993a. Reproductive systems and sibling competition in plants. Plant 588 Species Biology 8: 131–139. 589 CHEPLICK, G.P. 1993b. Sibling competition is a consequence of restricted dispersal in 590 an annual cleistogamous grass. Ecology 74: 2161–2164. 591 CHEPLICK, G.P. 1987. The ecology of amphicarpic plants. Trends in Ecology & 592 Evolution 2: 97–101. 593 CHEPTOU, P. O., O. CARRUE, S. ROUIFED, and A. CANTAREL. 2008. Rapid evolution of 594 seed dispersal in an urban environment in the weed Crepis sancta. Proceedings of 595 the National Academy of Sciences 105: 3796–3799. 596 CHEPTOU, P.O., J. LEPART, and J. ESCARRE. 2001. Differential outcrossing rates in 597 dispersing and non-dispersing achenes in the heterocarpic plant Crepis sancta 598 (Asteraceae). Evolutionary Ecology 15: 1–13. 599 CHEPTOU, P.O., and F. MASSOL. 2009. Pollination fluctuations drive evolutionary 600 syndromes linking dispersal and mating system. The American Naturalist 174: 46– 601 55. 602 CLAY, K. 1983. The differential establishment of seedlings from chasmogamous and 603 cleistogamous flowers in natural populations of the grass Danthonia spicata (L.) 604 Beauv. Oecologia 57: 183–188. 605 CODY, M.L., and J.M. OVERTON. 1996. Short-term evolution of reduced dispersal in 606 island plant populations. The Journal of Ecology 84: 53–61. 607 DARLING, E., K.E. SAMIS, and C.G. ECKERT. 2008. Increased seed dispersal potential 608 towards geographic range limits in a Pacific coast dune plant. New Phytologist 609 178: 424–435. 610 DE NETTANCOURT, D. 2001. Incompatibility and incongruity in wild and cultivated 611 plants. Springer-Verlag, New York, USA. 612 FRESNILLO, B., and B.K. EHLERS. 2008. Variation in dispersability among mainland and 613 island populations of three wind dispersed plant species. Plant Systematics and 614 Evolution 270: 243–255. 615 GIBSON, J.P. 2001. Ecological and genetic comparison between ray and disc achene 616 pools of the heteromorphic species Prionopsis ciliata (Asteraceae). International 617 Journal of Plant Sciences 162: 137–145. 618 GIBSON, J.P., AND TOMLINSON, A.D. 2002. Genetic diversity and mating system 619 comparisons between ray and disc achene seed pools of the heterocarpic species 620 Heterotheca subaxillaris (Asteraceae). International Journal of Plant Sciences 621 163: 1025–1034. 622 GUILLAUME, F., and N. PERRIN. 2006. Joint evolution of dispersal and inbreeding load. 623 Genetics 173: 497–509. 624 HARGREAVES, A. L., and ECKERT, C. G. 2014. Evolution of dispersal and mating systems 625 along geographic gradients: implications for shifting ranges. Functional 626 Ecology 28: 5-21. 627 HOLLAND, J.B., W.E. NYQUIST, and C.T. CERVANTES-MARTÍNEZ. 2003. Estimating and 628 interpretating heritability for plant breeding: an update. In J. Janick [ed.], Plant 629 breeding reviews vol. 22, 9–112. John Wiley & Sons, Inc., Hoboken, New Jersey, 630 USA. 631 HOTHORN, T., BRETZ F. and WESTFALL P. (2008). Simultaneous inference in general 632 parametric models. Biometrical Journal 50(3), 346-363. 633 IMBERT, E. 2001. Capitulum characters in a seed heteromorphic plant, Crepis sancta 634 25 (Asteraceae): variance partitioning and inference for the evolution of dispersal rate. 635 Heredity 86: 78–86. 636 IMBERT, E. 2002. Ecological consequences and ontogeny of seed heteromorphism. 637 Perspectives in Plant Ecology, Evolution and Systematics 5: 13–36. 638 IMBERT, E., ESCARRÉ, J. and LEPART, J. 1999. Differentiation among populations for 639 life-history, morphological and seed head traits in the seed heteromorphic species 640 Crepis sancta (L.) Bornm. (Asteraceae). – International Journal of Plant Sciences 641 160: 543–552. 642 IMBERT, E., and RONCE, O. 2001. Phenotypic plasticity for dispersal ability in the seed 643 heteromorphic Crepis sancta (Asteraceae). Oikos 93: 126–134. 644 IRITANI, R., and P.O. CHEPTOU. 2017. Joint evolution of differential seed dispersal and 645 self-fertilization. Journal of Evolutionary Ecology 30:1526-1543. 646 LENTH, R. (2018). emmeans: estimated marginal means, aka least-squares 647 means. R package version 1.2.2. https://CRAN.R-project.org/package=emmeans. 648 LLOYD, D.G. 1968. Variation strategies of plants in heterogeneous environments. 649 Biological Journal of the Linnaean Society 21: 357–385. 650 LYNCH, M., and J.B. WALSH. 1998. Genetics and analysis of quantitative traits. Sinauer, 651 Sunderland, MA. 652 MARSHALL, D.F., and R.J. ABBOTT. 1982. Polymorphism for outcrossing frequency at 653 the ray floret locus in Senecio vulgaris L. I. Evidence. Heredity 48: 227–235. 654 MARSHALL, D.F., and R.J. ABBOTT. 1984. Polymorphism for outcrossing frequency at 655 the ray floret locus in Senecio vulgaris L. II. Confirmation. Heredity 52: 331–336. 656 MARTÍN-FORÉS, I., ACOSTA-GALLO, B., CASTRO, I., DE MIGUEL, J. M, DEL POZO, A.M, 657 and CASADO, M. A. The invasiveness of Hypochaeris glabra (Asteraceae): 658 Responses in morphological and reproductive traits for exotic populations. PLoS 659 One. 2018; 13(6): e0198849. 660 MASSOL, F., and P.O. CHEPTOU. 2011a. Evolutionary syndromes linking dispersal and 661 mating system: the effect of autocorrelation in pollination conditions. Evolution 662 65: 591–598. 663 MASSOL, F., AND P.O. CHEPTOU. 2011b. When should we expect the evolutionary 664 association of self-fertilization and dispersal? Evolution 65: 1217–1220. 665 MCNAMARA, J., and J.A. QUINN. 1977. Resource allocation and reproduction in 666 populations of Amphicarpum purshii (Gramineae). American Journal of Botany 667 64: 17–23. 668 MIX, C., F.X. PICÓ, J.M. VAN GROENENDAEL, and N. JOOP OUBORG. 2006. Inbreeding 669 and soil conditions affect dispersal and components of performance of two plant 670 species in fragmented landscapes. Basic and Applied Ecology 7: 59–69. 671 OLIVIERI, I., Y. MICHALAKIS, and P.H. GOUYON. 1995. Metapopulation genetics and the 672 evolution of dispersal. The American Naturalist 146: 202–228. 673 OLIVIERI, I., M. SWAN, and P.H. GOUYON. 1983. Reproductive system and colonizing 674 strategy of two species of Carduus (Compositae). Oecologia 60: 114–117. 675 ORTIZ, M.A., S. TALAVERA, J.L. GARCIA-CASTANO, K. TREMETSBERGER, T. STUESSY, 676 F. BALAO, and R. CASIMIRO-SORIGUER. 2006. Self-incompatibility and floral 677 parameters in Hypochaeris sect. Hypochaeris (Asteraceae). American Journal of 678 Botany 93: 234–244. 679 ORTIZ-HERRERA, M. A., C. DE VEGA, S. TALAVERA. 2004. Hypochaeris salzmanniana 680 Fig. 3 Fig. 4