Saffron (Crocus sativus) is an autotriploid that evolved in Attica (Greece) from wild Crocus cartwrightianus

Crocus sativus is the source of saffron, which is made from dried stigmas of the plant. It is a male-sterile triploid that ever since its origin has been propagated vegetatively. The mode of evolution and area of origin of saffron are matters of long-lasting debates. Here we analyzed chloroplast genomes, genotyping-by-sequencing (GBS) data, nuclear single-copy genes, and genome sizes to solve these controversial issues. We could place 99.3% of saffron GBS alleles in Crocus cartwrightianus, a species occurring in southern mainland Greece and on Aegean islands, identifying it as the sole progenitor of saffron. Phylogenetic and population assignment analyses together with chloroplast polymorphisms indicated the wild C. cartwrightianus population south of Athens as most similar to C. sativus. We conclude that the crop is an autotriploid that evolved in Attica by combining two different genotypes of C. cartwrightianus. Vegetative propagation prevented afterwards segregation of the favorable traits of saffron.

individual that we found also in an individual of C. sativus. We used this marker as sequence !%(" characterized amplified region (SCAR) to screen for chloroplast differences in C. sativus and !%)" C. cartwrightianus individuals from all populations. We found that all C. sativus individuals !%*" possess the short allele while in C. cartwrightianus it occurs only in Attica and in one out of !&+" two populations from the island of Kea (Figure 2). Kea is directly adjacent of Attica (Figure

!&!"
1A) and was connected to the mainland repeatedly when Quaternary sea levels dropped !&#" (Lambeck, 1996). This may have allowed gene flow between the crocus stands of these

!&$"
areas. In all other populations of C. cartwrightianus only the longer allele was detected !&%"

!'("
To be able to infer the mode of origin of the triploid, i.e. if it evolved through a cross between !')" a di-and a tetraploid parent or through the combination of a reduced with an unreduced !'*" gamete within diploids, we collected leaves from 100 C. cartwrightianus individuals in the !(+" Attica area, dried them in the field in silica gel, and analyzed genome sizes for these !(!" individuals in the lab by flow cytometry. We were able to obtain results for 91 individuals. We

!(#"
found that all of them have a 2C genome size of 7.06 ±0.09 pg (SD). This value was also !($" observed in individuals with chromosome counts of 2n = 2x = 16. From this we conclude that !(%" C. cartwrightianus in Attica is diploid and that tetraploid plants are not frequent. This is in accord with the observation that C. sativus is a clonal lineage that originated most probably !('" only once, indicating that it was a rare event. If tetraploid individuals would occur regularly !(("

!)%"
Taking into account the high genetic diversity in C. cartwrightianus, we hypothesized that this

!)&"
could have influenced the outcome of earlier phylogenetic studies where C. cartwrightianus !)'" and C. sativus were included but did not result as sister species. We used DNA sequences !)(" of five single-copy genes that were amplified from C. cartwrightianus, C. sativus and their !))" four closest relatives C. hadriaticus, C. oreocreticus, C. pallasii and C. thomasii (Nemati et !)*" al., 2018), which all share the same chromosome base number of x = 8. Where initial direct !*+" sequencing provided evidence for the presence of more than one copy of a gene within an !*!" individual, amplicons were cloned and six clones per individual were sequenced.

!*#"
Phylogenetic analyses of the DNA sequences of the five genes ( Figure S5) revealed in all !*$" cases that alleles occurring in different species were not completely sorted according to their !*%" species affiliation. This phenomenon, referred to as incomplete lineage sorting (Maddison,

!*&"
1997), is often found among closely related species (Jakob and Blattner, 2006; Brassac and !*'" Blattner, 2015). Even the different alleles or homeologs detected within the C. sativus !*(" individual were rather diverse and could group in different clades in the gene trees. Thus, the !*)" high genetic diversity in C. cartwrightianus is not restricted to non-coding parts of the !**" genome (Larsen et al., 2015) but concerns also the gene space of this species, which could

#+$"
Earlier molecular studies of saffron evolution did not arrive at clear results regarding the #+%" parental species of C. sativus or the area of origin of saffron. The main reason seems that #+&" they did not take into account the high intra-specific genetic diversity present in C.

#+'"
cartwrightianus. Depending on the individual(s) studied and the marker region used, the #+(" resulting phylogenetic trees might reflect nearly arbitrary relationships ( Figure S5). In

#+)"
contrast, our GBS data were based on an exhaustive collection of C. cartwrightianus #+*" populations and clearly place the C. sativus individuals as sister of Attic C. cartwrightianus.

#!+"
Possible reasons for the sister group position instead of grouping within the Attic population #!!" is most probably the triploid and clonal nature of C. sativus that, as a group, has therefore a #!#" unique character combination that is in this way not present in any individual of C.

#!'"
When collecting leaves of Attic C. cartwrightianus, we evaluated the populations for the #!(" presence of the important traits typical for the saffron crocus. We recognized the bunchy #!)" growing habit, very long stigmas, stigmas of rather dark red color ( Figure 1D, Figure S6), and #!*" also stigmas with the specific taste and aroma of saffron, particularly in the southern part of ##+" this area. However, we did not find plants combining all these traits within single individuals ##!" in the same way as saffron. As genetic diversity is high in C. cartwrightianus and the species

###"
is an obligate outbreeder, it is unlikely to find in today's individuals regularly the exact allele ##$" combination characteristic for triploid C. sativus, as allele composition is constantly jumbled ##%" by genetic recombination. This is also apparent in the diverse karyotpyes of C.

##("
From Minoan frescos it is clear that more than 3600 years ago humans already used wild ##)" saffron in the southern Aegean. The first clear indication for the cultivation of triploid saffron

%#%"
We like to thank the Greek authorities for providing permits for plant collections, I.

%#&"
Faustmann, C. Koch, B. Kraenzlin and P. Oswald for help with plant cultivation and lab work,