Although outcrossing is the most prevalent mating system in angiosperms, a third of flowering plants exhibit mixed mating systems in which reproduction occurs both by self-fertilization and outcrossing (Goodwillie & al. 2005). In such cases, the genetic self-incompatibility system (SI) that is the main barrier to selfing in flowering plants (Igić & al. 2008) is eventually broken. In Asteraceae, while most species are self-incompatible (63%), a significant proportion are partially (10%) or totally (27%) self-compatible, pointing to a breakdown of their SI (Ferrer & Good-Avila 2007). Moreover, a complete and irreversible transition from outcrossing to selfing is a prevailing pattern in angiosperm diversification (Stebbins 1974; Igić & al. 2008; Goldberg & al. 2010). However, the maintenance of selfing requires being advantageous enough to counterbalance the effects of inbreeding depression (Charlesworth & Charlesworth 1987; Byers & Waller 1999). This situation may occur in populations with severe pollen limitation due to low effective population size and/or scarce pollinator presence (Busch & Schoen 2008; Good-Avila & al. 2008), in which selection for reproductive assurance is a prevalent factor (Busch & Delph 2012). This could be especially relevant in annual plants whose reproductive success is limited to one flowering season (Shivanna 2014, 2015).

The annual herb Anacyclus linearilobus Boiss. & Reut. is a narrow endemism occurring in coastal dune ecosystems in Algeria. This species was described in the surroundings of Oran by Boissier & Reuter (1852) and only few other populations are known so far (Humphries 1979). The genus Anacyclus was recently circumscribed to eight species distributed in western Mediterranean (Vitales & al. 2018; Álvarez 2019). Morphologically, A. linearilobus mainly differs from other Anacyclus species in the degree of basal leaves division, which in A. linearilobus is 1-2 pinnatisect whereas in the remaining species is tri-pinnatisect (Humphries 1979; Vitales & al. 2018). All species in this genus have a relatively wide distribution, except A. linearilobus that is restricted to the Algerian coast, and A. maroccanus Ball that is an endemism of the Morocco plateau (Humphries 1979; Vitales & al. 2018). Additionally, A. linearilobus is the only species in this genus growing in no anthropic habitats.

Anacyclus linearilobus was considered in the past closely related to A. clavatus (Desf.) Pers. and A. homogamos (Maire) Humphries (Humphries 1979), however recent phylogenetic analyses indicate A. valentinus L. as its sister species, in a clade including A. clavatus and A. radiatus Loisel. (Oberprieler 2004; Vitales & al. 2018). However, A. linearilobus and A. valentinus are morphologically quite different (i.e., while A. linearilobus show radiate capitula, A. valentinus present rayless ones). All species in Anacyclus are diploid (2n = 18; Humphries 1981; Rosato & al. 2017), although significant differences in genome size between them were documented (Agudo & al. 2019; Vitales & al. 2020). It is remarkable that the two extremes in genome size within the genus are represented by these two sister species, with sizes of 8.22 Gbp/2C in A. valentinus and 13.14 Gbp/2C in A. linearilobus (Vitales & al. 2020). Finally, it is important to consider the possible hybrid origin of A. valentinus suggested by Álvarez & al. (2020), since it could also involve its sister species, A. linearilobus. In hybrids, allelic incompatibilities (Bateson 1909; Dobzhansky 1936; Muller 1942) and lethal nuclear-cytoplasmic interactions (Levin 2003) may cause low seed sets and high variation in fertility.

The viability of interspecific crosses between all Anacyclus annual species has already been documented (Humphries 1981), but the breeding system has only been assessed in A. clavatus, A. homogamos, and A. valentinus (Álvarez & al. 2020). In all these cases, the species are considered self-incompatible and inter-fertile although reproductively isolated by postzygotic barriers.

Our goal here was to assess the breeding system in two populations of Anacyclus linearilobus and compare it with other annual species in the genus including its sister species, A. valentinus. We consider that this study is relevant not only for conservation purposes of A. linearilobus, but also for designing future research on the ecological and evolutionary singularity of this species.

Our results indicate that Anacyclus linearilobus has a mixed breeding system, in which most individuals are self-incompatible and coexist with other self-compatible and partially self-compatible ones. In contrast, its sister species, A. valentinus, as well as two other annual species in the genus, A. clavatus and A. homogamos, do not show a mixed reproductive system and are all considered self-incompatible (Álvarez & al. 2020). Therefore, it can be assumed that a breakage of the SI occurred along A. linearilobus evolution, allowing a shift to self-compatibility. The fact that the mating mixed system is present in the two populations studied and that transition to selfing is unidirectional (Igić & al. 2008) suggests that this reproductive strategy might be fixed at species level. However, occasionally, mixed mating systems may stabilize. For example, in the annual Hypochaeris salzmanniana DC. (Asteraceae), the proportion of selfing depends on pollinators availability (Arista & al. 2017), maintaining a mixed mating system over time. The paucity of pollinators might also be related with a partial shift to selfing in two isolated populations of the Madeiran endemism Tolpis succulenta (Aiton) Lowe (Asteraceae) (Crawford & al. 2019; Kerbs & al. 2020). The shift to self-compatibility may be associated with the floral biology, life cycle, and ecology of the species (Barrett 2014; Arista & al. 2017; Barrett & Harder 2017). Interestingly, unlike its sister species and others closely related, A. linearilobus is the only one growing in sand dunes, occupying a different habitat than its congeners. Besides, A. linearilobus individuals grow under shrubs canopy and therefore their occurrence is contingent to the shrub’s presence. While other Anacyclus species that occupy anthropic habitats may form dense continuous populations, in A. linearilobus the density of individuals might be lower with a more scattered distribution. Moreover, the presence and abundance of pollinators in these ecosystems might be heterogeneous and dependent as well on “vegetation islands” within the sand dune matrix. Under this situation pollen availability might be a limiting factor for outcrosses, while the effective population size decreases, driving to a need for a reproductive assurance strategy (Busch & Schoen 2008; Good-Avila & al. 2008), which is essential for species with short life cycles like annuals (Shivanna 2014). A similar scenario occurs in oceanic islands, in which selfing and mixed mating systems may be advantageous. In fact, in a study comparing species of Asteraceae, Brassicaceae and Solanaceae from islands and mainland, 66% of island species resulted self-compatible, compared to 41% of mainland ones (Grossenbacher & al. 2017).

The female reproductive success in Anacyclus linearilobus was variable both at individual and at population level (Fig. 2a, b). Similar type of variation has been observed at population level in A. valentinus (Fig. 2c, d), but not in other Anacyclus species studied (Fig. 2e-h). This variation in A. valentinus has been attributed to its possible hybrid origin (Álvarez & al. 2020). As A. linearilobus is the sister species of A. valentinus, we might consider a similar hybrid origin, and therefore, the variation observed in female reproductive success could also have similar causes. Other evidence supporting this hypothesis is the high variation in germination rates in A. linearilobus as well as in A. valentinus (Álvarez & al. 2020), and the difficulty in pollen release observed in some individuals of A. linearilobus, which is in agreement with the lower fitness expected in hybrids. Genic and allelic incompatibilities between pollen receptors and donors reduce mating availability, giving rise to a reduction in effective population size, which might be increased by inbreeding depression (Angeloni & al. 2011). In this case, it would be expected that seed set values for intra-population crosses were lower than for the inter-population ones. This was observed in the Bousfer population (Fig. 2a), but not in La Macta (Fig. 2b). As La Macta population has partially lost self-incompatibility, it is not expected that intra-population incompatibility would act as much as in the Bousfer population, resulting in a less variable and higher female reproductive success in La Macta (Fig. 2a, b). Other factors, such as species distribution, and population size and structure, have also relevant effects on reproductive performance. This seems evident in A. valentinus, which shows the highest variation in female reproductive success overall, but in which the seed set values for the intra-population crosses are significantly higher than for the inter-population ones (Fig. 2c, d). In this case, the large size of its populations occurring along roadsides, and its wide distribution in western Mediterranean coasts might have counterbalanced the effects of incompatibilities between individuals from the same population.

The assessment of breeding systems of rare or endemic species may be critical for the development of successful conservation strategies (Pérez & al. 2018). Since mixed mating systems may reduce species vulnerability (Yates & Ladd 2004), the finding that Anacyclus linearilobus shows this strategy as reproductive assurance is hopeful for the viability of its populations. However, the fact that this endemic species may experience high pollen limitation in the communities where it occurs (Alonso & al. 2010), its scattered distribution contingent to shrub canopy in the habitat it occupies, and its limitation in successful mating, make this endemic species a good candidate to be considered in conservation programmes. In addition, populations of A. linearilobus might be heavily impacted by development of industrial and tourism activities as well as the increase in the urban population in northern Algerian coastline (Snoussi & Aoul 2000). It might be possible that both the population of “Les Andalouses” near El-Ançor and that of the surroundings of Oran may have become extinct due to urbanistic development in these areas. Anacyclus linearilobus has a restricted distribution on the Algerian coast, with an extent of occurrence estimated as 4,227 km² and an area of occupancy estimated as 8-36 km² (of which only around 7 km² are suitable habitat). Therefore, we have suggested to the International Union for Conservation of Nature (IUCN) its assessment for the category of EN (Endangered).

In conclusion, in contrast with other species in Anacyclus that are considered self-incompatible, the narrow endemism A. linearilobus showed a mixed mating system in which most individuals were self-incompatible whereas others were partially self-incompatible, and others were self-compatible. Effective population sizes in A. linearilobus are diminished by intrinsic factors such the genic incompatibilities between different individuals giving rise to limitation in mating, but also maybe by environmental factors such as scattered habitat availability. Reproductive assurance under these circumstances may be favoured by the mixed mating system. Our results are also congruent with the hypothesis of a possible hybrid origin for A. linearilobus that showed a variable and relatively low female reproductive success both at population and individual levels. However, further research including genomic and cytogenetic data is required to be conclusive on the origin of this singular evolutionary entity. In the same line, the search for other populations of A. linearilobus along coastal sand dunes in Algeria and their demographic tracking would help in better assessing the state of conservation of this species.