Controlling for genetic identity of varieties, pollen contamination and stigma receptivity is essential to characterize the self‐incompatibility system of Olea europaea L.

Abstract Bervillé et al. express concern about the existence of the diallelic self‐incompatibility (DSI) system in Olea europaea, mainly because our model does not account for results from previous studies from their group that claimed to have documented asymmetry of the incompatibility response in reciprocal crosses. In this answer to their comment, we present original results based on reciprocal stigma tests that contradict conclusions from these studies. We show that, in our hands, not a single case of asymmetry was confirmed, endorsing that symmetry of incompatibility reactions seems to be the rule in Olive. We discuss three important aspects that were not taken into account in the studies cited in their comments and that can explain the discrepancy: (i) the vast uncertainty around the actual genetic identity of vernacular varieties, (ii) the risk of massive contamination associated with the pollination protocols that they used and (iii) the importance of checking for stigma receptivity in controlled crosses. These studies were thus poorly genetically controlled, and we stand by our original conclusion that Olive tree exhibits DSI.


| INTRODUCTION
Olive has been the iconic tree of the Mediterranean area due to its economical, ecological, cultural, and social importance over an extended period of human history, and a large number of studies have attempted to characterize its mating system. While no consensus model has emerged so far in the literature, some studies performed by the same group of authors (Breton & Bervillé, 2012;Breton et al., 2014;Farinelli, Breton, Famiani, & Bervillé, 2015;Koubouris, Breton, Metzidakis, & Vasilakakis, 2014) claimed to have identified at least six self-incompatibility alleles, and asymmetrical crosses indicative of a classical sporophytic self-incompatibility system. Our recent results (Saumitou-Laprade et al., 2017) contradict several of these previously published conclusions and indicate that Olive rather shares the unusual diallelic self-incompatibility (DSI) system previously discovered in Phillyrea angustifolia. Hence, according to those results, in Olive, only two incompatibility groups or genotypes do exist, with all individuals of a given group incompatible with each other and fully compatible with all individuals of the other group. As suggested in Saumitou-Laprade et al. (2017), this discrepancy is probably due to several shortcomings in previously published analysis that lacked proper genetic control. In our view, the Comment to the Editor by Breton et al. (2017) fails to take into account specific challenges associated with the genetic analysis of the particular biological material represented by Olive trees. Besides several inaccurate statements in their comment, we outline below three major sources of uncertainty in the studies cited by Breton et al. (2017) that prevent conclusive evidence to be drawn about the rejection of our model for selfincompatibility system of the Olive tree.
The first difficulty arises when comparing studies performed with plant material that is only referenced by variety's vernacular names.
Indeed, a major conclusion of our work (Saumitou-Laprade et al., 2017) was that it is essential to identify varieties by their reference genotype based on molecular markers rather than by their vernacular name (El Bakkali et al., 2013;Haouane et al., 2011;Trujillo et al., 2014), as this is associated with considerable confusion. Breton et al. (2017) contend that our results are inconsistent with their own studies. This is indeed true. To understand the origin of this discrepancy, we analyzed 66 trees from the olive collection cited in Saumitou-Laprade et al. (2017), which includes pro parte some of the varieties cited in the Breton et al. (2017) comment Farinelli et al., 2015). From these 66 individuals, we identified a total of 61 different genotypes using microsatellite markers and then matched these genotypes in the worldwide Olive World Germplasm Bank of INRA Marrakech assessed with the same markers. This simple analysis revealed no less than 14 cases where the genotypes associated with a given variety name were different among collections, representing a major discrepancy that demonstrates the unreliability of vernacular names. Hence, previous studies including those published by the Bervillé et al. group were based on poorly identified varieties, which is likely to have generated considerable uncertainty in the results. We believe that it will now be important for the community working on Olive trees to generate a public database of trees whose genetic identity has been ascertained by a common set of molecular markers, ideally also including their position in a reference orchard and a voucher DNA sample that could be exchanged among users. The Arabidopsis thaliana community has, for | 861

PERSPECTIVE
The second challenge is taking into account the risk of pollen contamination when performing controlled crosses in the Olive. A careful analysis of the methods typically used to perform controlled crosses in Olive reveals that, except in case of self-pollination in which the flowers remain protected during the whole process, the risk of contamination by pollen is indeed very high as Olive pollen is mostly wind dispersed. First, the twigs containing flowers to be pollinated are typically protected by a single bag in most studies. This protecting bag is opened at full blooming in the orchard to introduce pollen from fathers to be tested, either with a branch collected on the pollen donor tree or with a pencil. Massive contamination was demonstrated in Olea europaea in crosses following such a protocol (de la Rosa, James, & Tobutt, 2004) with as many as 96 of 149 (64%) of progenies whose expected father could be genetically excluded, and therefore resulting from pollen contamination. Similarly in Phillyrea angustifolia, a wild relative of Olive (Saumitou-Laprade et al., 2010), the paternity analysis of progenies produced by handmade, apparently controlled, crosses following a similar protocol, revealed more than 50% of progenies produced by contaminant pollen (unpublished data). Using a more carefully controlled pollination protocol (Billiard et al., 2015;Saumitou-Laprade et al., 2017), we showed that contamination could be decreased down to 1.7% (over 1,048 progenies tested, unpublished data). As we explained in Saumitou-Laprade et al. (2017), we strongly believe that any conclusion based on a protocol that entails such a massive level of contamination should be treated with caution, if not entirely disregarded, if it is not associated with molecular tests of paternity designed to exclude contaminated seeds.
Finally, according to the Breton et al. (2017) comment, the main objection against the existence of DSI in Olive remains its inability to explain the asymmetry reported by Breton et al. (2014) and other authors (Farinelli, Boco, & Tombesi, 2006;Farinelli et al., 2015;Moutier, 2006;Spinardi & Bassi, 2012;Villemur, Musho, Delmas, Maamar, & Ouksili, 1984) in studies based on measurement of fruit set following reciprocal pollination between pairs of varieties. Because the experimental protocol applied to assess the DSI in Olive (Saumitou-Laprade et al., 2017) was not designed to detect asymmetry in reciprocal crosses, we here present original results from reciprocal stigma tests performed with pairwise varieties for which asymmetry was published , strongly suggesting symmetrical instead of asymmetrical relationships. We also present results from diallelic crossing experiments performed with eight different varieties which could explain why so many symmetrical crosses between compatible varieties have been interpreted as asymmetrical.

| PLANT MATERIAL AND METHODS TO ASSESS ASYMMETRY IN RECIPROCAL CROSSES
We worked with some varieties cited in Breton et al. (2014). We Stigma and pollen were collected on each individual tree minimizing the risk of pollen contamination, each individual tree was phenotyped for SI and was genotyped by 15 polymorphic microsatellite marker loci (Baldoni et al., 2009;El Bakkali et al., 2013). Therefore, we provide for each individual: a SI phenotype, a physical position in the orchard, a genotype corresponding to a specific combination of alleles at 15 polymorphic SSR loci (see Table S1 in Saumitou-Laprade et al. (2017) for the 16 genotypes shared with the previous study, and Table S1 in the present study for Oit46).
In June 2013 and 2014, 17 different genotypes were chosen in the orchard and assigned to one of the two SI groups using stigma tests, as described in Saumitou-Laprade et al. (2017). In a first experiment, nine genotypes were selected in order to replicate nine pairwise compatibility tests between varieties for which asymmetry has been reported in Breton et al. (2014). Reciprocal cross-compatibility was assessed using stigma tests in pairwise tests (see Table 1 and Figure 1 (Table   S1), we used 10-days-old stigma. Stigma were protected from contaminant pollen by double bagging and transferred to the laboratory in bags still closed 10 days after the first flower opened on the tree, harvested from twigs under laboratory conditions, maintained 24 hr in petri dishes containing a Brewbaker and Kwack medium (Vernet et al., 2016) and pollinated.

| Symmetrical rather than asymmetrical incompatibility reactions in Olive
In eight cases, we observed pollen tubes converging through the stigmatic tissue toward the style until the base of the stigma and entrance of the transmitting tissue of the style, indicating perfect compatibility between parents of the crosses (see Table 1 and Figure 1, panels 1-8).
In all reciprocal crosses, compatibility was observed in both directions of the reciprocal crosses. In the last cross, we observed only short pollen tubes that did not reach the style (see Table 1 (2017). Pairs 1-8 correspond to compatible crosses (conclusion = 1) among varieties belonging to two different SI groups; pair 9 corresponds to incompatible cross (conclusion = 0) among varieties belonging to the same group. Phenotyped trees are labeled according to their reference genotype (Saumitou-Laprade et al., 2017) and their variety name in the studied orchard. nd: not defined incompatibility relationships detected are all in agreement with the SI group assignment performed previously with stigma tests (see Table   S1 in Saumitou-Laprade et al. (2017)) or during the current study (  Breton et al. (2014) were accompanied by paternity analyses, pollen contamination may have produced unreliable results. Nevertheless, it is worth mentioning that pollen contamination can only explain "false positive" errors: that is, seeds produced by means of crosses that should otherwise be incompatible (for instance, the seeds expected on "Rosciola" when pollinated by "Carolea": See fig. 2, scheme F in Breton et al. (2014)).
To explain the absence of seeds produced by one of the two compatible parents in the eight additional crosses, that is, "false-negative" errors, we analyzed the diallelic scheme among eight different Olive varieties (Table 2) In red, dicrepancy detected between predicted and observed compatibility/incompatibility relationships among varieties.
be fertilized by pollen from the later. We suggest that asymmetries reported in literature may correspond to false-negative results between compatible mates whose periods of blooming were not sufficiently synchronized.

| CONCLUSIONS
The mating system of the Olive tree has remained a controversial issue in the literature, but many of the previously published studies have been based on a poorly genetically controlled experimental design.
Given 1) the vast uncertainty around the genetic identity of vernacular varieties, 2) the massive risk of contamination associated with commonly used pollination protocols and 3) the importance of checking  (Bergelson et al., 2016;Griffith, Owens, & Thuman, 2002). We believe that would be now a good time for the Olive tree research community to join this general movement. The cross described in Table 2 is correctly described in Material and Methods section, but there is a mistake in the legend title of the table: It is written (Oit64 × Oit27) instead of (Oit27 × Oit15).

| FEW REMARKS IN RESPONSE TO
We apologize for this mistake and thank the authors of the comment for their remark. Nevertheless, this error does not change the conclusions of the genetic analysis of the cross which shows the 1:1 segregation of progenies that are all self-incompatible and equally distributed among the two SI groups (and not a "segregation for self-fertility" as written in the comment).

Arguments in favor of the sporophytic nature of the SI in O. europaea.
We underlined in Saumitou-Laprade et al. (2017) that none of the arguments presented in literature was decisive and we presented two arguments based on original results we obtained. The first argument that establishes the sporophytic nature of the self-incompatibility system refers to the 1:1 proportion of the two parental SI groups in the controlled-cross progeny that excludes the possibility of gametophytic control of self-incompatibility (GSI) (Bateman, 1952). The second argument refers to the requirement of GSI, to be functional, of a minimum of three S alleles (with strict codominance between S alleles in the pistil to avoid compatibility of heterozygous individuals), and that defines a minimum of three incompatibility groups (Hiscock & McInnis, 2003). The two groups observed in O. europaea can be explained, only by a sporophytic diallelic SI system.

DATA AND MATERIAL SHARING
All relevant data are within the paper and its Supporting Information files.