1. Introduction
Polyploidy is one of the most important and ubiquitous driving forces in plant evolution [
1,
2,
3]. In a climate-change context, Levin (2019) proposed that polyploidization will be one of the most frequent speciation modes in the next 500 years, with an increase in the percentage of polyploid plants from 35% today to more than 50%. This hypothesis is related to the fact that polyploidy can occur in sympatry, and neopolyploids can be established in only a few generations; therefore, they might be reinforced in rapidly changing scenarios [
4].
Autopolyploids are considered to form by genome multiplication, while allopolyploids derive from hybridization between species with differentiated genomes, either by chromosome doubling after the fusion of reduced or unreduced gametes. However, this classification has long since been discussed. The term “segmental allopolyploid” is used to describe polyploids that do not exhibit strict bivalent formation across all chromosomes or disomic inheritance at all loci [
5]. While autopolyploids and allopolyploids are identified according to chromosome pairing behavior (formation of multivalents and bivalents in meiosis, respectively), Stebbins (1947) considered that the parents of segmental allopolyploids occupied an intermediate level of chromosomal divergence between those of autopolyploids and allopolyploids [
6].
How polyploids establish remains a debated question in evolutionary biology [
7], especially autopolyploids that seem to have no clear evolutionary advantage over their diploid progenitors [
8,
9]. As a result, for years, autopolyploids were considered rare in nature, representing evolutionary dead ends [
10,
11]. However, recent studies showed that autopolyploids are more abundant than expected [
12,
13], and their abundance could have been underestimated due to recognition difficulties, as their phenotypes are similar or identical to their diploid progenitors [
14].
To analyze the evolutionary significance of polyploidy, Levy and Feldman (2004) differentiated short-term “revolutionary changes”, related mainly to the establishment of polyploids, and long-term “evolutionary changes”, related more to their expansion and persistence. In this sense, allopolyploids undergo extensive genomic changes in first generations [
15,
16,
17], while autopolyploids may experience fewer structural changes [
13,
18]. Unlike nascent autopolyploids, well-established autopolyploids can also show substantial genome reorganization compared to their diploid relatives [
8,
13,
19]. Along with these genomic changes, functional reorganization of the gene-expression network may also occur, which is much more evident in allopolyploids than in autopolyploids [
20,
21,
22]. Recent studies considered autopolyploidy as a macromutation with epigenetic consequences [
23]. In general, hybridization (included in allopolyploid formation) seems to trigger significant changes, while only genome doubling maintains a similar state to that of its diploid progenitor [
6,
8,
24]. In addition, autopolyploids have always been expected to be less fertile than allopolyploids are given the meiotic irregularities caused by multivalent chromosome pairing [
25]. However, more recent studies revealed that aberrant meiosis affects both auto- and allopolyploids [
17,
26]. Such irregularities may be overcome through the frequent turnover of reproduction modes (from sexual to apomictic reproduction) [
27,
28] or the evolution of a stabilized form of meiotic asymmetry in chromosome inheritance [
29].
The union of unreduced (2n) gametes is thought to be the commonest pathway for natural polyploid formation [
30] through either the fusion of two unreduced gametes or a “triploid bridge” that can generate tetraploid progeny through selfing or backcrossing [
31]. Triploids are often sterile [
32,
33], but they can sometimes produce a large proportion of fertile unreduced gametes that increases the possibility of tetraploid formation [
31,
34]. The production of unreduced gametes has been proven heritable, governed by a few genes, and increasing with environmental stress, such as heat, frost, water deficit, and herbivory [
8,
35,
36]. This is in accordance with the higher frequency of polyploids found in habitats affected by climate fluctuations and disruptions [
37,
38,
39]. Some studies pointed out that polyploid species are over-represented in previously glaciated regions, while diploids are more frequent in disjunct refugial areas [
40,
41]. Under these abiotic conditions, polyploid formation can be recurrent and with multiple origins. Despite all this knowledge, the establishment of neopolyploids and especially neo-auto-polyploids remains unclear.
Centaurea (Asteraceae) is a taxonomically intricate genus due to the existence of polyploidy, descending dysploidy cycles, and hybridization events, with a large number of polyploid complexes [
42,
43]. The
Centaurea aspera L. polyploid complex has long since been studied, and it is mainly distributed in coastal habitats of Spain and Morocco. [
32,
38,
44,
45,
46,
47,
48,
49,
50,
51]. What makes the
C. aspera polyploid complex so interesting is that it is made up of natural populations of the parental diploid (
C. aspera L. 2x = 22), an allotetraploid (
C. seridis L. 4x = 44) and an autotetraploid (
C. gentilii Braun-Blanq. and Maire 4x = 44). Chromosome counts were previously performed by cytological techniques in the three studied species [
49].
Centaurea seridis and
C. aspera can coexist in Spanish natural-contact zones to produce triploid hybrids (
C. × subdecurrens Pau 3x = 33) [
32]. Similarly,
C. gentilii and
C. seridis can coexist in Morocco to produce tetraploid hybrids (
C. × paucispina (Ferriol, Merle and Garmendia) P.P. Ferrer 4x = 44) [
46,
49]. The existence of these taxa with different ploidy levels, geographical distributions, fertility, and mating systems allows for direct comparison to analyze their competitive advantages or disadvantages.
In this context, we aimed to study the mating system and reproductive barriers among Centaurea aspera and its polyploid relatives, allotetraploid C. seridis (4x = 44), and autotetraploid C. gentilii (4x = 44), and to analyze their geographical distribution. We specifically addressed the following questions: (i) what are the geographical distribution and ploidy level of the Moroccan populations? (ii) Do the seed sets that derive from the intra-specific crosses within each taxon differ? (iii) Is hybridization between C. aspera and C. gentilii possible? (iv) Are the seed sets per capitulum that derive from intra- and inter-specific crosses different? We combined previously published information with the new data to offer an overview of the geographical distribution and reproductive behavior of the three species and their hybrids.
4. Materials and Methods
4.1. Geographical Distribution and Ploidy Level of Moroccan Populations
The geographical distribution, ploidy level, and genetic analyses of the Spanish populations of
C. aspera and
C. seridis were already reported [
47,
48]. Moroccan populations were surveyed for the first time in 2013. Then, eight populations of
C. gentilii were located on the Atlantic coast of Morocco [
49]. During two new expeditions (2016 and 2017), the Atlantic and Mediterranean coasts of Morocco were exhaustively resurveyed from Tiznit (southern point) to Nador (northeastern point). To determine the populations’ ploidy level, individuals were sampled during the 2016 expedition, 309 from four
C. gentilii populations, and 96 from the
C. seridis populations (
Table 6). Thirty
C. aspera individuals from the Spanish populations were resampled to confirm their ploidy level. The ploidy level of these individuals was determined by flow cytometry as described by Garmendia et al. (2015).
4.2. Controlled Pollinations
During the 2017 expedition, cypselae were collected for forced-pollination experiments. Cypselae were sampled from two natural populations of each species (
Table 1). The sampled capitula from the natural populations were stored at 4 °C for 2 months. In all cases, the sampled capitula came from 4–5 mothers under open-pollination conditions. One hundred cypselae were randomly extracted from the mixture of capitula sampled from each population and germinated. Individuals of the three species were grown in the Centro para la Investigación y la Experimentación Forestal (CIEF; Quart de Poblet, Spain) greenhouse. At least 50 plants from each species and population were grown in pots for the experiments done in January 2018.
Four controlled pollination, treatments were performed in these plants: intra-specific crosses within the three species, and inter-specific crosses between C. aspera and C. gentilii. The treated capitula were randomly selected from those available to obtain at least 30 treated capitula per treatment and taxon. Treatments were run during the flowering period, from June to August 2018. Pollinations were performed with the newly open capitula bagged in semipermeable nylon bags prior to anthesis. Upon anthesis, capitula were brushed gently against one another once a day on 2 consecutive days. During the cross-pollinations, each treated capitulum received pollen from the one-paired capitula from a different individual. The flowers of the treated capitula were not emasculated.
Due to plant management, the Zaouiat individuals grew late and bloomed outside the 2018 season, which was why the inter-specific crosses between C. aspera and C. gentilii from Zaouiat were repeated in 2019. Additionally, in 2019, the intra-specific crosses in C. aspera and C. gentilii (both allogamous) were specifically repeated to compare the seed sets obtained using individuals from the same population or from different populations within the taxon. For C. aspera, four pollination treatments were performed: (ss), ovules and pollen from El Saler; (sc), ovules from El Saler and pollen from Chulilla; (cs), ovules from Chulilla and pollen from El Saler; (cc), ovules and pollen from Chulilla. For C. gentilii, four cross-treatments were also performed: (tt), ovules and pollen from Tamri; (tz), ovules from Tamri and pollen from Zaouiat; (zt), ovules from Zaouiat and pollen from Tamri; (zz), ovules and pollen from Zaouiat.
4.3. Progeny Analysis
After pollinations, capitula were rebagged for 6 weeks until fruit set. For each treatment, total cypselae per capitulum were counted. Cypselae were disinfected with 0.5% NaClO solution for 20 min, washed 3 times for 5 min in distilled water, and hydrated on parafilm-closed Petri dishes for 24 h at 20 °C. Subsequently, each cypsela was cut at 2/3 from the epicotyl, and the pericarp was removed. Both empty (without embryo) and intact (with fully developed embryos) cypselae were counted. In the intact cypselae, the 1/3 distal cotyledonary tissue was used to determine the ploidy level of each embryo by flow cytometry, as described by Garmendia et al. (2015). Each sample consisted of a small piece of leaf (0.5 cm2) collected from the plant, to be analyzed together with a similar leaf piece taken from a diploid control plant. Samples were chopped together using a razor blade in a nucleus isolation solution (High-Resolution DNA Kit Type P, solution A; Sysmex Partec, Munster, Germany). Nuclei were filtered through a 30 µm nylon filter and stained with a 4,6-diamine-2-phenylindol (DAPI) solution (High-Resolution DNA Kit Type P, solution B; Partec). After a 5 min incubation period, stained samples were run in a CyFlow Ploidy Analyzer (Partec) flow cytometer equipped with optical parameters for the detection of DAPI fluorescence. DNA fluorochrome DAPI was excited with UV–LED at 365 nm. Histograms were analyzed with CyView software (Sysmex Partec, Munster, Germany), which determines sample peak position, coefficient of variation (CV), arithmetic mean, and median. The rest of the embryo (2/3) was placed on wet Petri dishes at room temperature and in natural light so they could germinate.
4.4. Statistical Analyses
The average, standard error, skewness, and kurtosis of the number of cypselae per capitulum were assessed for each pollination treatment and taxon. The normality of residuals and homogeneity of variances were checked by a Shapiro–Wilk test and Levene test, respectively. Due to the lack of normality for the
C. aspera and
C. gentilii residuals, nonparametric methods were selected to compare medians: the median number of cypselae was compared among treatments and repetitions by the Kruskal–Wallis rank sum [
69] and post hoc Dunn’s [
70] tests. Pearson residuals were used to highlight the significant differences between observed/expected frequencies of full/empty cypselae. All statistical analyses, tables, and figures were constructed using Stat graphics XVII-X64 and R language [
71] with RStudio [
72].