Natural interploidy hybridization among the key taxa involved in the origin of horticultural chrysanthemums

Understanding hybridization and introgression between natural plant populations can give important insights into the origins of cultivated species. Recent studies suggest differences in ploidy may not create such strong reproductive barriers as once thought, and thus studies into cultivated origins should examine all co-occurring taxa, including those with contrasting ploidy levels. Here, we characterized hybridization between Chrysanthemum indicum, Chrysanthemum vestitum and Chrysanthemum vestitum var. latifolium, the most important wild species involved in the origins of cultivated chrysanthemums. We analysed population structure of 317 Chrysanthemum accessions based on 13 microsatellite markers and sequenced chloroplast trnL-trnF for a subset of 103 Chrysanthemum accessions. We identified three distinct genetic clusters, corresponding to the three taxa. We detected 20 hybrids between species of different ploidy levels, of which 19 were between C. indicum (4x) and C. vestitum (6x) and one was between C. indicum and C. vestitum var. latifolium (6x). Fourteen hybrids between C. indicum and C. vestitum were from one of the five study sites. Chrysanthemum vestitum and C. vestitum var. latifolium share only one chloroplast haplotype. The substantially different number of hybrids between hybridizing species was likely due to different levels of reproductive isolation coupled with environmental selection against hybrids. In addition, human activities may play a role in the different patterns of hybridization among populations.

7 quality of isolated genomic DNA was assessed on 1.0 % agarose gels, and then 151 diluted to a concentration of 10-20 ng/ul for genotyping and sequencing. Thirteen 152 microsatellite loci were used for genotyping (Zhang et al., 2014;Jo et al., 2015). The

Population genetic analysis
157 It is often difficult to assign microsatellite genotypes for mixed ploidy species, as the 158 frequency of different alleles can be difficult to quantify. In hexaploids, each 159 microsatellite locus would be expected to have up to six alleles per individual. We 160 chose to score each allele separately using the software GENEMARKER 2.4.0 161 (Softgenetics), and checked each genotype manually. We then calculated allele 162 richness for each population using FSTAT 2.9.4 (Goudet, 1995) and performed 163 principal coordinate (PCO) analysis in POLYSAT 1.7-4 (Clark and Jasieniuk, 2011), 164 based on Bruvo's pairwise genetic distances (Bruvo et al., 2004). 165 We performed STRUCTURE analysis for each hybridizing population separately 166 using STRUCTURE 2. 3.4 (Pritchard et al., 2000) with ploidy specified as 6n. We 167 combined XG, NX and PH into one hybridizing population as the three localities are 168 separated by only a few kilometers. We set the number of genetic clusters (K) to 2 169 when analyzing the genetic structure of each hybridizing population as only two 170 parental species are present. The allopatric C. indicum populations (TA and ZP) were 171 used as a reference population.

172
In addition, to identify the most likely K value across populations we included all 173 populations in a combined STRUCTURE analysis, testing K values from 1 to 10. The 174 number of genetic clusters was estimated using the Evanno test (Evanno et al., 2005) 8 in the program Structure Harvester 0.6.94 (Earl and vonHoldt, 2012). Ten replicates 176 of the STRUCTURE analysis were performed with 1,000,000 iterations and a burn-in 177 of 100,000 for each run. The admixture model, with an assumption of correlated allele 178 frequencies, was used. Individuals were assigned to clusters based on the highest 179 membership coefficient averaged over the ten independent runs. Replicate runs were 180 grouped based on a symmetrical similarity coefficient of >0.9 using the Greedy using the Kruskal-Wallis test in the R package agricolae v1.3-3 (de Mendiburu, 2020). 195 We would expect that the average allele number is higher for C. vestitum and C.

Hybridization across ploidy levels inferred from microsatellites 227
Genetic diversity estimates were similar among the three taxa and hybrids, with allelic 228 richness ranging from 3.74 to 4.26 and gene diversity from 0.75 to 0.84 (Table S3).

229
On average, the number of alleles scored in C. vestitum and C. vestitum var. latifolium 230 was significantly higher than C. indicum at eight and seven loci, respectively (P < 231 0.05); this was expected for a hexaploid possessing more chromosome copies than a 232 tetraploid species (Fig. S1).

299
Consistent with this, we detected 20 hybrids among C. indicum, C. vestitum and C.
hybrids is likely to be underestimated in our study, because of the stringent 302 confidence threshold applied in the STRUCTURE analysis. Three individuals with 303 admixture between 26.2%-31.4 from population XG and two individuals with 304 admixture between 34.5%-48.7% were not supported by Q scores with 95% CIs, but 305 may prove to be hybrids such as later generation backcrosses. hybrids may fail to survive due to the breakdown of suites of co-adapted genes.