The experiments were conducted on a walnut collection belonging to the Research Station of Department of Horticultural Sciences, University of Tehran and on walnut plants from private orchards ( Table 2 ). The followings were used as plant materials for molecular progeny tests: the progeny plants of two commercial walnut genotypes (P7-M/GBP, P8-M/Pedro) whose pistils and catkins were completely destroyed by freezing event and who set seeds from the secondary blooming without developing catkins, and the progeny of six walnut genotypes (P1-M/G52_DB_41, P2-M/G44_ZT, P3-M/Jamal, P4-M/G55_PDB, P5-M/G11_1, P6-M/G51_DB) able to set seeds after destruction of catkins by frosts and whose pistillate flowers were isolated by bags (bags made of Span band polypropylene (PP) polymers).

Five walnut genotypes were used for cytohistological investigation ( Table 3 ). Isolated female flowers were collected every other day up to two months after anthesis (soon after they were distinguishable from the leaves rosette), followed by sampling every four days up to one month later if sufficient seeds were available. All the collected materials were fixed in FAA solution including ethanol (96%): formaldehyde: glacial acetic acid (17:2:1) and stored in ethanol (70%), followed by dehydration in an ascendant ethanol series (80, 90, 95,100%), clearing by immersing in different proportions of absolute ethanol: toluene (2:1, 1:1, 1:2) and toluene only (repeated twice), and finally embedded in paraffin and sectioned at 7 µm slices with a rotary Micro DC 4055 microtome (Dideh Sabz, Urmia, Iran). After deparaffinization by toluene and a descendant ethanol series (100, 90, 80, 70, 50%), specimens were immersed in distilled water, stained using Meyer’s hematoxylin techniques and contrasted by eosin. Serial sections were studied under a light microscope (LABOMED model LX50 connected to LABOMED digital camera, iVu 3100) and a fluorescence microscope (BELL connected to BELL digital camera, Model BLACKL 3000, Italy). The developmental stages of the ovule, including megasporogenesis, megagametogenesis, and embryo development were evaluated from the perspective of reproductive strategy in the laboratory of Plant Cell Developmental Biology, Department of Biology, Bu-Ali Sina University, Hamedan, Iran. At least 50 flowers were analyzed for each stage listed above.

Genomic DNA was extracted from 200 mg of freeze-dried fresh leaves per sample using a modified CTAB method reported by de la Cruz et al. (1997) and Sánchez-Hernández and Gaytán-Oyarzún (2006). Thirty microsatellite regions or SSR loci of walnut were chosen from the literature (Woeste et al., 2002). The amplification efficiency and the polymorphism rate of the 30 SSR primer pairs were initially examined in singleplex and multiplex reactions on two samples chosen based on high phenotypic diversity in the laboratory of Genomics for Plant Breeding, DAFNAE, University of Padova, Italy.

The 7 SSR loci ( Table 4 ) that amplified efficiently, providing unequivocal and polymorphic profiles, were then organized in two multiplexes and validated on 110 samples including the maternal parents (n=8) and their offspring (n=102; Table 2 ). The forward primers were tagged by the addition of oligo sequence tails M13, PAN1, PAN2 and PAN3 labeled to the 5’ end to be used in PCR along with the complementary fluorophore oligonucleotides 6-FAM, NED, PET and VIC. The PCR reaction consisted of diluted gDNA template (20 ng μL-1), Platinum Multiplex PCR Master Mix (1x) (Applied Biosystems, Carlsbad, CA, USA), 10% GC enhancer (Applied Biosystems), 0.05 μM tailed forward primer (Invitrogen, Carlsbad, CA, USA), 0.1 μM reverse primer (Invitrogen), 0.23 μM universal primer (Invitrogen) and sterile dd-water to reach a reaction volume of 20 μL. A 9600 Thermal Cycler (Applied Biosystems) was used with the following conditions: 5 min at 95°C, followed by 5 cycles of 95°C for 30 s, 54/58°C (T_a belonging to two different multiplexes) for 45 s, with the temperature being decreased by 1°C with each cycle, and 72°C for 45 s; then 35 cycles of 95°C for 30 s, 49/53°C (T_a belonging to two different multiplexes) for 45 s, and 72°C for 45 s, followed by the final extension step of 60°C for 30 min. Amplicons were first run by means of agarose gel electrophoresis (2% agarose): TAE buffer (1x): Sybr Safe DNA stain (1x) (Life Technologies) and visualized on an Uvidoc HD6 transilluminator (Uvitec, Cambridge, UK) equipped with a digital camera. Amplification products were then dried at 65°C for 1 h and subjected to capillary electrophoresis (ABI PRISM 3130xl Genetic Analyzer, Thermo Fisher), and the alleles were sized using Peak Scanner Software 1.0 (Applied Biosystems) with LIZ 500 (Applied Biosystems) as an internal size standard.

Measures of genetic diversity, including the total number of observed alleles (N_a), the number of effective alleles (N_e), the observed and expected heterozygosity (Ho/He), and the Shannon index (I) of phenotypic diversity for each locus, were estimated using GenAlEx 6.5 software. The polymorphism information content (PIC) was calculated using Excel Microsatellite Toolkit as follows:

In which n is the total number of alleles detected for the locus and p_i is the frequency of the ith allele of the locus in the set of 110 samples investigated (Anderson et al., 1993).

Pairwise genetic similarity estimates were calculated between maternal plants and their progeny plants in all possible comparisons based on the Jaccard similarity index (Jaccard, 1908). Hierarchical clustering (UPGMA) with the probability value AU (approximately unbiased p value) of multiscale bootstrap resampling method (bootstrap replicates=1000) was obtained through the function implemented in the “pvclust” (Suzuki and Shimodaira, 2006) package of the free R software (version 4.2.2).

Ploidy level was evaluated by flow cytometry analysis of stained nuclei by 4′,6-diamidino-2-phenylindole (DAPI), isolated from leaves of 40 walnut progenies in the laboratory of Genomics for Plant Breeding, DAFNAE, University of Padova, Italy. In brief, 35 progenies that -based on SSR data- were predicted to be the result of outcrossing were analyzed to identify any possible polyploid BIII hybrid. Similarly, 5 progenies that -based on SSR data - were predicted to be the result of selfing were analyzed to identify any possible parthenogenesis-derived haploid accession through flow cytometry (CyFlow Ploidy Analyzer, Sysmex, DE). Following the procedure outlined in the CyStain UV Precise protocol, Approximately, 0.5 cm² of young fresh leaves were chopped with 0.5 ml of Nuclei Extraction Buffer (Sysmex Partec) and incubated for 2 minutes at room temperature. Then, each sample was filtered (30 µm CellTrics®, Sysmex) and incubated in 2 ml of staining buffer for 60 sec before analysis (blue fluorescence emission= 435-500 nm; flow rate of 4 µl/sec). Fluorescence histograms were evaluated using Flowing software 2.

We described and quantified the developmental stages of the walnut ovule in five different maternal walnut genotypes, detailing megasporogenesis and megagametogenesis to detect the predominant reproductive strategy ( Table 3 ). It is necessary to specify that among the examined genotypes, micrographs were taken from two of them (i.e. P1-M/G52_DB-41 and P2-M/G44_ZT) for summarizing the main events and features of female sporogenesis and gametogenesis.

The walnut ovary was found to contain a single orthotropous ovule, which is unitegmic, crassinucellate and derived from the basal placenta ( Figure 1 ). All genotypes demonstrated identical patterns during the megasporogenesis stages. A distinct megasporocyte or megaspore mother cell (MMC) was found to be situated roughly in the middle of at least five to six parietal layers that ultimately separate the MMC from the nucellar epidermis ( Figures 1A, F ). This cell progressed through the first meiosis, forming a dyad, and the second meiosis, yielding a linear tetrad of megaspores ( Figure 1 ), which morphologically resembled those found in the sexual process. At this point, the functional megaspore at the chalazal end was already more extensive and more vacuolated than the others, and evolved into the embryo sac. The other three micropylar megaspores naturally degenerated ( Figures 1C, I, J ). MMCs typically underwent complete meiosis, except in some cases (0.8-1%; Table 3 ), in which abnormalities were found, including more than one linear tetrad ( Figure 2A ), followed by two embryo sacs (4%; Table 3 ) lying side by side or end to end ( Figures 2B, C ). Additionally, (2%; Table 3 ), binuclear functional megaspores ( Figures 1D, E ) were rarely observed instead of normal uninuclear products, possibly indicating post-meiotic restitution or endoreduplication.

We found no significant differences between the genotypes during the early and final mega-gametogenesis phases. Three consecutive mitotic divisions progressed through a functional megaspore to generate an eight-nucleate embryo sac. A two-nucleate embryo sac was formed as a consequence of the first mitosis, in which two nuclei migrated to the opposite poles of the micropylar-chalazal axis ( Figures 3A, B, E, F ). These two nuclei were then divided into a four-nucleate embryo sac ( Figures 3C, G ). Eight nuclei ( Figures 3D, H ), four at the chalazal pole and four at the micropylar pole, resulted after the third and final mitosis. All genotypes demonstrated the polygonum type (seven-celled/eight-nucleate) of the embryo sac ( Table 3 ), recognized by three antipodals, two polar nuclei, and an egg apparatus (two synergids and the egg cell) in the final phase of mega-gametogenesis ( Figure 4 ). Both embryo and endosperm tissues followed developmental stages typical of fertilized ovaries after double fertilization. Once the embryo reached the 32 or 64-cell stage, the cellular endosperm constitution was quite evident at micropylar regions ( Figure 5B ). Further embryo development stages including, 8-cell, 64-cell, globular, heart-shaped, torpedo-shaped, and cotyledon stage ( Figures 5A–E ) conformed the zygotic nature of the embryos located in the expected region, close to micropylar end ( Figures 5F, G ). Nearly 17% of the embryos were found positioned far from the expected micropylar end zone, at the middle of embryo sac ( Figures 5H–J ; Table 5 ). Furthermore, we observed that very few ovules contained an embryo-like structure outside the embryo sac (adventitious embryony) and or polyembryony in genotype P1-M/G52_DB-41 ( Figures 5I, J ).

Based on cytological data available in the literature, both diplosporous apomixis and adventitious embryony seems to be possible in J. regia. Despite Sartorius and Stösser (1997) hypothesized that walnut could be characterized by diplospory, our cytohistological results did not support the occurrence of apomeiosis. However, we observed few post-meiotic anomalies in megasporogenesis and, in particular, the formation of binuclear functional megaspores in genotype P1-M/G52_DB_41. This was previously observed in Miconia albicans after meiosis II, supporting the hypothesis of endoreduplication or restitution events (Caetano et al., 2013). However, in our genotypes the MMC typically went through complete meiosis and the ovules displayed the typical polygonum type development of the embryo sac. As a matter of fact, neither diplosporic mechanisms nor aposporic initials have been documented. Our results indicated that embryo development started with egg cell division inside the embryo sac. About 80% of embryos were positioned close to the micropylar end zone inside the embryo sac. The observation of a few embryos positioned far from micropylar end with unknown origin and one embryo outside the embryo sac were in agreement with Schanderl (1964) and Valdiviesso (1990) were the apomixis-derived embryo resulted located apart from micropylar end from nucellus somatic cells, as a result of adventitious embryony.

We used 7 polymorphic SSR nuclear loci to genotype 102 progenies obtained from controlled crosses (isolated by bagging) and after a second pistillate flowers production (after freezing events). The related mother plants (n=8) were also genotyped ( Table S1 ). Overall, thirty-five marker alleles, ranging from 4 (WGA 4, and WGA 42) to 7 alleles (WGA 65) per locus, were observed among the 110 walnut individuals with a mean value of 5 alleles per locus ( Table 6 ). This value was in full agreement with what observed by Ebrahimi et al. (2011) (i.e. 5.1), but lower than those reported by Aradhya et al. (2010), and Ebrahimi et al. (2017) (10 and 7, respectively), possibly due to the small number of mother plants initially used in the experiment. The number of effective alleles ranged from 2.11 (WGA 42) to 5.26 (WGA 65). Based on the Botstein et al. (1980) classification, all the SSRs used were highly polymorphic, with PIC values always exceeding 0.5. The mean value of the polymorphism information content (PIC) across all marker loci was 0.61, with estimates ranging from a minimum of 0.52 (WGA 42) to a maximum of 0.81 (WGA 65). In addition, the mean value of Shannon’s information index (I) was 1.12, ranging from 0.82 (WGA 42) to 1.74 (WGA 65) ( Table 6 ), lower than the values reported by Vahdati et al. (2015) (I = 1.79). The analysis of plant genotypes revealed an average observed heterozygosity (Ho) equal to 0.65, varying from 0.35 to 0.95. Overall, the average heterozygosity for single individuals was found to be relatively high, consistently with the putatively high outcrossing rates of walnut.

All maternal genotypes generated unique molecular profiles based on the seven marker loci ( Table S1 ) and resulted therefore distinguishable from each other. The mother plants were heterozygous for at least 3 (up to 6) microsatellite loci, which was sufficient to distinguish different origins of reproduction. Each progeny plant was analyzed for all seven SSR loci and a UPGMA tree was then constructed for each population using the matrix of the genetic similarity estimates calculated in all possible pairwise combinations ( Figure 6 ). Based on the genotyping data, three possible scenarios were expected: i) offspring with a molecular profile identical to that of the mother plant, likely resulting from apomixis ( Figure 7A ); ii) offspring with a molecular profile characterized by maternal and non-maternal (paternal) alleles, and therefore the result of outcrossing ( Figure 7A ); iii) offspring with a molecular profile characterized by some of the possible maternal alleles (and no non-maternal alleles), likely resulting from selfing ( Figure 7B ).

Considering all 7 SSR loci, none of the offspring was genetically identical to their mother parent and hence apomixis was not recorded in the investigated genotypes ( Figure 6 and Table 7 ).

Populations 2, 5, 6, 7 and 8 resulted exclusively constituted by progenies deriving from outcrossing. Indeed, each of the above-mentioned offspring showed non-maternal alleles in at least one SSR locus. In contrast, the proportion of progenies originated by selfing in populations 1, 3 and 4 ranged between 17% and 33% ( Table 7 ). The similarity coefficient of individual progenies originated by outcrossing when compared to their mother parent ranged from a minimum of 0.71 to a maximum of 0.95. By contrast, the genetic similarity values calculated between progenies originated by selfing and their mother parent ranged from 0.88 to 0.97 ( Table 7 and Figure 8 ).

All the samples (n=40) analyzed by flow cytometry resulted diploids ( Figure 9 ). For 33 out of 40 samples, this finding was in agreement with the SSR data. On the contrary, the remaining seven samples were expected to be triploids, in light of the WGA 65 triallelic SSR locus. This result partially confirms the hypothesis according to which a single three-allelic locus does not always imply triploidy (Draga et al., 2023). Furthermore, since no haploid or polyploid sample was observed, the cytometric analysis was also useful to demonstrate the total lack of accessions classifiable as non-hybrid or BIII hybrid.

Apomixis rates (defined as fruit set rate after bagging isolation) of 9.5-15.6% (Zhang et al., 2000), 13.1-76.8% (Guo-Liang et al., 2007) and 24.7% (Guoliang et al., 2010) have been reported for walnut trees grown in China. On the other hand, in the Romanian and Turkish walnut cultivars studied by Cosmulescu et al. (2012) and Şan and Dumanoğlu (2006), data showed that the seed set rate after bagging was low and insufficient for commercial seed and crop production. Bagging isolation has been so far the most commonly employed method for investigating apomixis in walnut. However, in most of the previous studies, the twigs containing flowers were typically protected by a single bag and, in species were the pollen is mostly wind dispersed (including walnut), the risk of contamination by pollen remains very high (Saomitou-Laprade et al., 2017). In this study, despite isolating female flowers by bagging and considering that the male catkins of the genotypes studied were mainly destroyed by the spring frost, we proved that 92% of progeny plants were derived from outcrossing, and 8% were produced by selfing. This would suggest that bagging isolation alone cannot be relied upon to determine whether apomixis occurs in walnut, and that complementary molecular progeny tests and cytological analyses of the progenies are needed. Our genotyping analyses did not support any asexual mode of reproduction and all individuals studied were identified as zygotic plants. Based on cytometric analysis, all progenies resulting from outcrossing and selfing were diploids, indicating that triploid and haploid status was not prevalent among the tested genotypes. On the other hand, based on cytohistological results, gametophytic apomixis is unlikely to occur in the walnut genotypes studied, although adventitious embryony cannot be completely ruled out, as the embryos derived from sporophytic apomixis are commonly diploid.

The present findings resulted in agreement with general expectations considering previous reports in walnut (Wood, 1934; Loiko, 1990). According to these studies, some pistils are produced at second or third time upon new flash growth following the freeze event, apparently without any possible chance of receiving pollen or developing nuts. In the absence of further complementary evidence, it has so far been hypothesized that a possible progeny may have been formed via apomixis (even with rates up to 78.8% - 81.2%). On the contrary, combining cytohistological, genotypic and flow citometry analyses, we were able to prove that all the progenies originated through a sexual process and with a high outcrossing rate, despite the use of bagged mother plants whose primary pistils and catkins were completely destroyed by freezing weather, and whose secondary blooming did not involve catkins development.