An efficient method to produce segregating populations in quinoa (
 Chenopodium quinoa
 )

Rapid global population growth together with the lack of availability of arable land and accessible water is imposing a challenge especially in many poor countries, to feed the population with sufficient and nutritious food. The pseudocereal quinoa, a member of the Amaranthaceae family, has gained increasing interest as an alternative staple food particularly in marginal lands, due to its high nutritional value and strong tolerance to abiotic stresses like drought, salinity, frost and heat (Jacobsen et al., 2003). Quinoa originated and has been cultivated along the Andes region of South America for the last 7,000 years (Williams & Brenner, 1995). Quinoa seeds have a high protein level (10 to 18%) with a perfect balance of amino acids specifically the essential amino acids (Vilche et al., 2003). They also provide a valuable combination of beneficial micronutrients like potassium, copper, zinc, iron and calcium along with fibre, lipids, carbohydrates and vitamins (Vega-Gálvez et al., 2010). Moreover quinoa is low in gluten, which offers a perfect substitute for wheat for people suffering from celiac disease. Despite of its exceptional characteristics, the seed yield of quinoa is generally low (around 1–2 t/ha), which makes breeding activities inevitable. Activities to breed varieties with higher seed yield Received: 2 June 2020 | Revised: 10 July 2020 | Accepted: 28 September 2020 DOI: 10.1111/pbr.12873

| 1191 EMRANI Et Al. and good nutritional quality were driven from a small number of accessions in the Altiplano, which constitutes a very narrow genetic base for quinoa breeding programmes (Jacobsen & Mujica, 2002). Therefore, along with the conservation of landraces, which is essential for the preservation of the genetic material, efforts should be concentrated on the introduction of new germplasm into breeding programmes to increase genetic diversity. Quinoa is mainly an autogamous species with very small flowers, which particularly complicates crosses. Therefore, development of an efficient crossing method should serve as the first step in quinoa breeding programmes.
Quinoa produces panicle inflorescences consisting of mostly hermaphrodite, but also pistillate flowers. The hermaphrodite flowers consist of five sepals and five stamens surrounding the ovary, and two stigmas (Abdelbar, 2018;Figure 1a). The hand emasculation of quinoa flowers has shown to be very difficult due to the floral morphology, the size of the flowers and the rapid progress of flowering within the inflorescence (Jacobsen & Stølen, 1993;Peterson et al., 2015). In many species, heat treatment and vacuum emasculation have been routinely applied as alternative methods for physical emasculation of the seed parents (Almeida et al., 2017;Mukasa et al., 2007;Otsuka et al., 2010;Sha, 2013). Warm water emasculation relies on the fact that pollen is generally more sensitive to higher temperatures than the ovary and the stigma. Therefore, selecting the appropriate temperature, at which the pollen will not be viable anymore, but the ovary and stigma are still active (usually around 45°C), is crucial for the application of this method (Sha, 2013). Almeida et al. (2017) reported the hyperthermotherapy in a water bath of 46°C for a panicle immersion time of 2.5 min to be more effective for the production of sterile plants compared to vacuum emasculation in rice. However, warm water emasculation has been reported as ineffective in quinoa, because it damages the inflorescence (Fleming & Galwey, 1995). Stetter et al. (2016) considered three different crossing methods (open pollination, warm water emasculation and hand emasculation) for creating inter-and intra-specific hybrids in three grain species of the genus Amaranth. Their results indicated hand emasculation and open pollination to be the most and the least efficient methods for creating hybrids, respectively.
As an alternative to mechanical emasculation, male sterility systems can be considered as the method of choice, particularly for the production of hybrid varieties in a commercial level. Different sources for male sterility have been identified in quinoa germplasm (Simmonds, 1971;Ward & Johnson, 1994), including a cytoplasmic male sterility system, for which restorer genes can be found in many quinoa accessions (Ward, 1998). However, introducing this CMS system into parents to facilitate crossings is time-consuming.
Moreover using the CMS system in quinoa breeding programmes does not seem to be feasible in the near future, since heterosis has not been reported yet and therefore commercial hybrid production seems elusive at the moment. Breeding inbred line varieties is still the method of choice as the floral morphology of quinoa rather promotes line development.
Apart from the availability of efficient methods for crossing, strategies for the identification of F 1 plants are also required to produce a large array of F 2 populations. Two strategies for the identification of F 1 plants have been proposed using morphological and/ or molecular markers. In case of the former, parents should differ by easy to detect qualitative morphological traits with dominance/ recessive inheritance. In quinoa, seed colour, inflorescence color, axillary pigmentation, and plant colour could be used as morphological markers, if the pollinator is homozygous for the dominant alleles (Peterson et al., 2015). This approach would however limit F I G U R E 1 Floral morphology, crossing methods and selection of hybrid plants. (a) A hermaphrodite flower during pollination (stage BBCH60). The stamen consists of five anthers and the filaments form a ring around the ovary of the carpel. (b) Warm water emasculation crossing method. (c) Growing seed parent and pollinator side by side under a plastic bag ('no emasculation method'). (d) Axil pigmentation as morphological marker for selection of F 1 plants. Seed parents with green axil pigmentation were crossed with red-axil pigmented pollinators. The F 1 plants showed red axil pigmentation. Photos were taken at BBCH59 the number of crosses that can be performed, because only parents with contrasting phenotypes can be used. On the contrary, selection with molecular markers would allow crosses between any two parents, even if they show exactly the same morphological traits.
Theoretically, one marker that is polymorphic between the two parents would be enough for the selection of true F 1 plants.
In the current study, we compared three different crossing methods to produce F 1 seeds. We found that one polymorphic marker locus unequivocally distinguishes F 1 hybrid plants from selfing offspring. In combination with phenotypic selection, we propose a two-step selection procedure for the identification of true F 1 plants.
Moreover we produced 30 different F 2 populations, which can serve as starting material in quinoa breeding programmes.

| Plant material and growth conditions
To test different crossing methods and their efficiency in quinoa (Chenopodium quinoa Willd.), we considered 12 red-axil and 11 green-axil quinoa accessions from five different countries of origin for crossing (Table 1). Accessions with the dominant morphological trait (red axil) were considered as male parent and accessions with the recessive morphological trait (green axil) were considered as female parents for the crossing experiment (Table 1). We planted two plants per accession in 13 cm 2 pots in a cold greenhouse in May 2018 under natural long-day conditions in Kiel, Germany. To synchronize the flowering time of both crossing parents, we sowed the late-flowering accessions in the first week and then considered five different sowing dates in five consecutive weeks for the early-flowering accessions, based on the flowering time data of all the accessions from previous experiments (data not shown). We harvested the seeds on the seed parent and sowed 280 seeds per cross in 35x-multi trays in September 2018 in the cold greenhouse to identify the hybrid progenies. After leaf sampling for DNA isolation, we transferred the putative F 1 plants to 13 cm 2 pots for efficient seed production. All the true F 1 plants were bag-isolated to produce F 2 seeds.

| Crossing methods
We considered three different crossing methods for the production of F 1 seeds (Figure 1b,c and Figure 2).

| Hand emasculation of the seed parent
For this method, we followed the procedure suggested by Peterson et al. (2015) with slight modifications. To reduce the number of flower clusters to a manageable number, the flower bud of the seed parent was removed with scissors, once it was visible on the top of the plant (approximately 1.5 cm in size). We kept only 3-4 flower clusters on the seed parent. Then we opened the flowers under a magnifying lens (with 10x magnification) and removed all five anthers with a tweezer, when they were still green or yellowish-green ( Figure 2).

| Emasculation of the seed parent with warm water
The inflorescence of the seed parents was dipped into a water bath of 45°C for 10 min (Figure 1b). This procedure was previously reported to be successful for the emasculation of the seed parents in grain amaranth (Stetter et al., 2016).

| No emasculation
Here we did not perform any treatments for the emasculation of the seed parent and simply placed the seed parent and the pollinator together under an isolation bag (Figure 1c).

| DNA extraction
We collected leaf samples from all the putative F 1 plants and lyophilized them before DNA isolation. We used the NucleoSpin® Plant II DNA isolation kit (Macherey-Nagel) or CTAB method with slight modifications (Saghai-Maroof et al., 1984) to extract the DNA from the dried leaf samples.

| Confirmation of crosses
We conducted a two-stage selection for the identification of F 1 plants. First, we selected all the putative F 1 plants based on axil pigmentation. We phenotyped the seedlings for axil pigmentation four weeks after sowing. The putative F 1 plants are the ones that show red axil pigmentation, although their seeds have been collected on a green-axil seed parent ( Figure 1d). We expect that these plants are hybrid plants, since they show the dominant phenotype they have inherited from the pollinator. Additionally, we considered at least one green-axil plant from each cross as control for genotyping.
As a second step, we used the publicly available Insertion-Deletion (InDel) markers described in Zhang et al. (2017) to confirm the genotype of the F 1 plants (Table S1). The parental lines of each cross were first screened with the InDel markers to find the polymorphic markers for each cross combination. These markers were then used for the identification of true F 1 plants from the putative ones ( Figure 3).
Every PCR reaction with the InDel markers had a total volume of 20 µl with the following amplification conditions: 94°C for 5 min as initial denaturation, 35 cycles of: 30 s at 94°C, 30 s at primer pair annealing temperature and 30 s at 72°C and a final extension step of 10 min at 72°C.

| Success rate evaluation and statistical analysis
For the calculation of success rates and selection accuracy and for their statistical analysis, we used the statistical software R (Team, R. C, 2019). We used appropriate generalized linear models (McCullagh & Nelder, 1989) with the parental accessions combination and crossing method as influence factors. No interaction effects were assumed.
The residuals were assumed to follow a binomial distribution. For success rates, the distribution is based on the number of confirmed F 1 plants and the total number of the plants phenotyped for each cross.
For selection accuracy, the distribution is based on the number of confirmed F 1 plants using molecular markers and the total number of putative F 1 plants for a given cross, which were selected using the phenotypic selection method. In this context and in order to enhance the model estimates, we used a Bayesian generalized linear model described by Gelman et al. (2008) for data analysis. Based on the model for the success rates, an analysis of variances was conducted, followed by multiple contrast tests (Bretz et al., 2011) in order to (a) identify the superior parental combinations and (b) compare the crossing methods.

| RE SULTS
We produced in total 214 F 1 hybrids from 30 crosses between 23 parents using three crossing methods. All the hybrid plants were bag-isolated and selfed to produce 30 F 2 populations. The crossing parents were selected based on their axil pigmentation which had been determined before (Table 1). We considered three crossing methods (no emasculation, hand emasculation and warm water emasculation) and the parental combinations as sources of variation.
Our results showed that both parental combinations and crossing methods have a significant effect on the success rate of the crossing experiment (ρ = <2.20E-16 and 2.32E-10, respectively). We used a two-step method for the identification of F 1 plants from all the crosses. As axil pigmentation has been reported to have a dominant monogenic inheritance (Simmonds, 1971), we expect all the F 1 plants from a cross between a green-axil seed parent and a red-axil pollinator to have red axils ( Figure 1d). Therefore, we only considered genotyping with the molecular markers for the red-axil plants together with a green-axil plant as control. For genotyping, we used six publicly available InDel markers (Zhang et al., 2017), which were polymorphic between the parents ( Not<del author="Nazgol Emrani" command="Delete" timestamp="1602617894146" title="Deleted by Nazgol Emrani on 10/13/2020, 9:38:14 PM" class="reU3">e</del>: For the identification of hybrid plants, genotyping with the published InDel markers was considered (Zhang et al., 2017). The size of the amplicons in base pairs is written in brackets for seed parent and pollinator, respectively. Success rates were calculated based on the number of confirmed F 1 plants and the total number of the plants phenotyped for each cross, while selection accuracy rate was calculated based on the number of confirmed F 1 plants using molecular markers and the total number of putative F 1 plants for a given cross, which were selected using axil pigmentation.
comparable to the accuracy rate we reported for axil pigmentation, considering the low number of investigated crosses for hypocotyl colour.
The comparison of different crossing methods revealed significant differences between all three methods ( produced the least number of F 1 plants with an average success rate of 0.06% (Table 2). To select the most successful parental combinations for the production of hybrids, we identified the crosses, which showed a higher success rate over all three crossing methods compared to the overall mean of the whole experiment. Five crossing combinations produced hybrid seeds with a significantly higher success rate compared to the mean success rate of the experiment (Table 4).
We wanted to know if the phylogenetic relationship of the parents plays a role in the success rate of their cross. Therefore, we used 1.6 million SNPs derived from the whole genome re-sequencing of 20 out of the 23 investigated accessions in this study for phylogenetic analysis. The individual's dissimilarity and coancestry coefficient grouped the accessions in two main clusters (group I and group II; Figure 4). Nine out of the ten (90%) accessions in group I were originated from Peru or Bolivia, while all but one of the Chilean accessions investigated in this study were grouped in group II. Group I represents quinoa highland accessions in our experiment, while accessions in group II belong to the coastal quinoa population (Jarvis et al., 2017). We found crosses between more related accessions  Note: For the statistical analysis, the logit estimates of the success rates were calculated. *, **, ***: parental genotype combinations that show a higher success rate compared to the mean of all the crosses at α = 0.1, α = 0.05, α = 0.01 and α = 0.001, respectively. The ρ-values are based on appropriate multiple contrast tests (Bretz et al., 2011). it is crucial to develop a simple method for crossing and hybrid seed detection. Manual emasculation has been suggested as a promising method for production of hybrid seeds. However, hand emasculation is reported to be very difficult in quinoa due to its compact inflorescence and its tiny florets (Jacobsen & Stølen, 1993), while warm water emasculation has been considered ineffective due to the damage to the inflorescence (Fleming & Galwey, 1995). Despite of the challenges of the hand emasculation of quinoa flowers, attempts have been made to develop F 2 populations after hand crossing in quinoa. A detailed instruction for hand emasculation of quinoa flowers under controlled conditions is available (Peterson et al., 2015). Nonetheless, this protocol only describes the crossing procedure and does not report on further selection of F 1 plants using molecular markers. Here, we present a two-step selection strategy for the selection of true F 1 plants based on morphological and molecular markers. In this way, we created 30 segregating populations from crosses between closely related as well as distantly related quinoa accessions.

| D ISCUSS I ON
We reasoned that two components (parental combination and crossing method) would account for the crossing success. Quinoa is an autogamous species with highly variable outcrossing rates (0.5 to 17%) across different accessions (Murphy et al., 2016). In a crossing experiment, parental accessions with a high outcrossing rate can be Quinoa accessions, which produce less cleistogamous and more chasmogamous flowers with exposed anthers and abundant pollen over a longer period of time, would be ideal candidates as pollinators in crossing programmes. Additionally, accessions with higher number of pistillate flowers would be more suitable as seed parents in crossing programmes. We would suggest phenotyping these traits in different quinoa accessions to identify putative candidates as crossing parents for breeding programmes in quinoa.
Apart from mating type and floral morphology, pollinators with dominant qualitative morphological traits can facilitate the identification of F 1 plants. In the current experiment, we selected different parents from highland and coastal origin displaying high genetic diversity (Jarvis et al., 2017). To reduce the molecular marker screening workload, we considered axil pigmentation for the first selection step of the F 1 plants. It is important to mention, that the identification of red axil plants is not always easy. In some cases, axils of putative F 1 plants display a pink colour, which makes it hard to distinguish them from the green axils of selfing progenies. Moreover axil pigmentation may appear sporadically, which makes it difficult to detect (Peterson et al., 2015). Therefore, the selection accuracy due to axil pigmentation was highly variable and not suitable as a sole criterium to detect hybrids. There are other morphological traits like stem colour, inflorescence colour, saponin content or seed colour, which can be considered for the selection of F 1 plants. In our study, the parental combination had a significant effect on the success rate of the cross. We identified five parental combinations, which produced significantly higher numbers of hybrid plants. By comparing the most successful crosses, it is evident that the accessions 171,024 and 170,916 were present in more than one significantly successful cross as seed parent and pollinator, respectively. Among the five most successful crosses, two crosses were made between highland ecotypes, while one cross was made between the coastal accessions and two crosses were made between different ecotype groups. Therefore, considering the material explored in the current study, we did not see any associations between the phylogenetic relationship of the accessions and the success rate of the crosses. This suggests that crossing success in quinoa will be independent of the parental origin. It is noteworthy that hand emasculation has been successfully used to generate interspecific hybrids between C. quinoa and C. berlandieri, which opens new perspectives for breeding quinoa for disease resistances (Bonifacio, 2003).
Hand emasculation was the most successful crossing method in the current study, followed by warm water emasculation and no emasculation, which was in line with the previous report in grain amaranth (Stetter et al., 2016). However, we recorded a lower success rate for our most efficient method (maximum success rate of 55.94% for hand emasculation) compared to the reported success rates in amaranth (74% for hand emasculation; Stetter et al., 2016). We believe, this difference is due to slight differences in the execution of the crossing methods. In the previous study in amaranth, the emasculation was repeated after seven days of the first emasculation and any flowers developed after the emasculation were removed. In our study due to technical reasons and the high number of crosses that we performed, repeated emasculation of the seed parents and removal of extra flowers were not technically feasible. Therefore, we expected a higher percentage of selfing seeds in the crossing progenies, which was a reason to consider 280 seeds per crossing event for identification of F 1 plants.
In future experiments, we will consider multiple emasculation events and elimination of the intact flowers on the seed parents, to increase the success rate of the crosses. Nevertheless, it is evident that the hand emasculation significantly increases the success rate of the crosses and should be considered to assure the production of hybrid seeds. could significantly shorten the growth cycle of amaranth plants using the same strategy. A speed breeding protocol for shortening the growth cycle of quinoa is already available (Ghosh et al., 2018).

Differences in
However, this protocol is only designed for day-neutral/long-day quinoa accessions. Using the measures mentioned above, we expect to be able to shorten the growth cycle of short-day quinoa accessions, which encompass the majority of quinoa germplasm.
We produced 30 segregating populations from crosses between genetically diverse quinoa accessions, which can serve as starting material for breeding programmes in quinoa. The improved progenies of the crosses can be selected and investigated in consecutive generations using the established breeding methods like pedigree or single seed descent. Moreover the F 2 populations and their progenies can be used for QTL mapping and identification of candidate genes for agronomically important traits in quinoa.

ACK N OWLED G EM ENTS
The authors thank Monika Bruisch for her support in conducting the experiment and performing the crosses in the greenhouse. The authors also acknowledge Verena Kowalewski and Brigitte Neidhardt-Olf for their support in the lab. Additionally, the authors also thank Prof. Mark Tester (King Abdullah University of Science and Technology, Saudi Arabia) for providing seeds for some of the parental accessions.

CO N FLI C T O F I NTE R E S T
The authors declare that there is no competing interest.

AUTH O R CO NTR I B UTI O N S
NE designed the experiment, performed the crosses and wrote the manuscript. MH performed the statistical analysis. DS calculated the phylogenetic relationship between the parental accessions. NM performed genotyping of the progenies using the molecular markers.
ER produced the SNP file for the parental accessions. CJ supervised the study and critically revised the manuscript. All authors read and approved the manuscript.