Evidence of Genomic Exchanges between Homeologous Chromosomes in a Cross of Peanut with Newly Synthetized Allotetraploid Hybrids

Cultivated peanut and synthetics are allotetraploids (2n = 4x = 40) with two homeologous sets of chromosomes. Meiosis in allotetraploid peanut is generally thought to show diploid-like behavior. However, a recent study pointed out the occurrence of recombination between homeologous chromosomes, especially when synthetic allotetraploids are used, challenging the view of disomic inheritance in peanut. In this study, we investigated the meiotic behavior of allotetraploid peanut using 380 SSR markers and 90 F2 progeny derived from the cross between Arachis hypogaea cv Fleur 11 (AABB) and ISATGR278-18 (AAKK), a synthetic allotetraploid that harbors a K-genome that was reported to pair with the cultivated B-genome during meiosis. Segregation analysis of SSR markers showed 42 codominant SSRs with unexpected null bands among some progeny. Chi-square tests for these loci deviate from the expected 1:2:1 Mendelian ratio under disomic inheritance. A linkage map of 357 codominant loci aligned on 20 linkage groups (LGs) with a total length of 1728 cM, averaging 5.1 cM between markers, was developed. Among the 10 homeologous sets of LGs, one set consisted of markers that all segregated in a polysomic-like pattern, six in a likely disomic pattern and the three remaining in a mixed pattern with disomic and polysomic loci clustered on the same LG. Moreover, we reported a substitution of homeologous chromosomes in some progeny. Our results suggest that the homeologous recombination events occurred between the A and K genomes in the newly synthesized allotetraploid and have been highlighted in the progeny. Homeologous exchanges are rarely observed in tetraploid peanut and have not yet been reported for AAKK and AABB genomes. The implications of these results on peanut breeding are discussed.


INTRODUCTION
Allopolyploids and autopolyploids are two different types of polyploids, each resulting from a different genetic origin and showing a distinct meiotic behavior (Xu et al., 2015). In autopolyploids, all chromosome sets are identical or very closely related, while allopolyploids have divergent chromosome sets (Leitch and Leitch, 2008;Soltis et al., 2014). Sets of homologous chromosomes are considered "homeologous" to other sets from the other genomes (Sybenga, 1975, Sybenga, 1996. Thus, in classic autopolyploids, each chromosome may pair randomly with any of its homologs in equal frequencies during meiosis (Muller, 1914), leading to a polysomic inheritance. Autopolyploids can undergo double reduction where the segments of two sister chromatids end up in the same gamete (Haynes and Douches, 1993;Mather, 1936). In contrast to the situation in autopolyploids, the cytological diploidization of allopolyploids requires a nonrandom assortment of chromosomes into pairs, of which crossovers are exclusively formed between homologous chromosomes with disomic inheritance at each locus (Cifuentes et al., 2010;Gaeta and Chris Pires, 2010). However, in addition to these extremes, an intermediate pattern of genetic inheritance has also been described (Stebbins, 1947;Sybenga, 1996;Stift et al., 2008;Jeridi et al., 2012).
Cultivated peanut, Arachis hypogaea L., is one of the major oilseeds and cash crops worldwide for which genetic improvement can tremendously benefit from its wild relatives (Rami et al., 2014). This species is autogamous and allotetraploid (2n = 4x = 40), harboring homeologous A and B genomes (Husted, 1936;Smartt et al., 1978). It is assumed that it originated from a single hybridization event between two wild diploid taxa , most likely Arachis duranensis (A genome) and Arachis ipaensis (B genome), followed by a spontaneous chromosome duplication (Seijo et al., 2004(Seijo et al., , 2007Bertioli et al., 2016). The single origin of the crop, superimposed with domestication, resulted in a severe genetic bottleneck. Therefore, closely related diploid wild species, which have maintained a high genetic diversity, were considered suitable to broaden the genetic basis of the cultivated gene pool (Simpson, 2001;Favero et al., 2006;Foncéka et al., 2009). The wild relatives of cultivated peanut are mostly diploid (2n = 2x = 20) and contain species with A, B, D, K, and F genomes (Smartt et al., 1978;Stalker, 1991;Robledo and Seijo, 2010).
Given the ploidy difference between the wild and cultivated peanut, the production of colchicine-induced allotetraploids was used as a pathway to introduce wild alleles into the cultivated gene pool (Simpson, 1991). Several synthetic allotetraploids that have been produced by crossing different diploid species have proven to be cross-fertile with A. hypogaea (Mallikarjuna et al., 2011). Moreover, the development of peanut genomics tools has made possible the marker-assisted introgression of wild genes into a cultivated background (Fonceka et al., 2012).
However, this breeding approach has raised new fundamental questions on the meiotic behavior of the synthetic allotetraploid used in breeding programs and the possible genetic changes related to their genomic composition. Meiotic instabilities are common in interspecific and resynthesized lines (Gaeta et al., 2007;Lyrene, 2016). In some allopolyploid plants, recombination between subgenomes during meiosis was suspected to occur in newly formed polyploids (Ramsey and Schemske, 2002;Soltis et al., 2010) but was rarely observed among stabilized allopolyploids (Salmon et al., 2010;Ainouche and Wendel, 2014).
Based on the classic genetic behavior of the allotetraploid genome and cytogenetic observations, several genetic mapping studies in peanut have been conducted considering a diploidlike behavior at meiosis (Burow et al., 2001;Hong et al., 2008Hong et al., , 2010Varshney et al., 2008;Foncéka et al., 2009;Qin et al., 2011;Shirasawa et al., 2013;Zhou et al., 2014). However, recently, thanks to a thorough analysis of genotyping data, Leal-Bertioli et al. (2015) reported unexpected missing and rare single nucleotide polymorphism (SNP) genotypes in recombinant inbred lines derived from a cross between a cultivated peanut and a synthetic allotetraploid. The authors showed that these missing data could be explained by the occurrence of partial tetrasomic recombination. Recombination among homeologous chromosomes is poorly understood in tetraploid peanut and the exact types of meiotic behavior remain unclear although the determination of these factors is important for our knowledge and for the development of appropriate breeding strategies. In addition, the classical impact of non-disomic inheritance on the genomic structure, such as segregation distortion and double reduction have not yet been reported.
In this study, we investigated the meiotic behavior of tetraploid peanut based on the segregation patterns of 380 microsatellites markers in F 2 progeny derived from the cross between the cultivated peanut Arachis hypogaea and a synthetic allotetraploid (Arachis duranensis × Arachis batizocoi) 4x . To obtain more insight into the mode of inheritance at the genome scale, we analyzed recombination events between homologous and homeologous chromosomes in relation to their position on a genetic linkage map. We reported the occurrence of a mixture of disomic and polysomic modes of inheritance of SSR loci, confirming the recent partial tetrasomic assumption made by Leal-Bertioli et al. (2015). We showed that this mixed inheritance was consistently associated with segregation distortion and homeologous chromosome substitution in some progeny. Our results suggest that the homeologous recombination events occurred between the A and K genomes in the newly synthesized allotetraploid and have been highlighted in the progeny.

Plant Material and Population Development
The study was conducted using an F 2 population derived from the cross between the cultivated variety Fleur 11 and the synthetic allotetraploid ISATGR 278-18 (Arachis duranensis × Arachis batizocoi) 4x . Fleur 11 is a Spanish type with erect growth habit widely cultivated in West Africa. ISATGR 278-18 was developed and kindly provided by ICRISAT-India (Mallikarjuna et al., 2011). The synthetic allotetraploid combines the AA genome of A. duranensis (ICG 8138; 2n = 2x = 20), a close wild relative and one of the most probable ancestors of A. hypogaea, and the KK genome of A. batizocoï (ICG 13160; 2n = 2x = 20), a wild relative taxa that was reported to pair with the B genome of the cultivated species during meiosis (Burow et al., 2001;Mallikarjuna et al., 2011;Leal-Bertioli et al., 2014).
ISATGR 278-18 was reported to have a normal chromosome configuration with 20 bivalents (Mallikarjuna et al., 2011). However, several cycles of self-pollination were performed at CERAAS prior to hybridization with Fleur 11. Five plants of the synthetic allotetraploid were used as male to cross with five plants of Fleur 11 used as female. The F 1 plants were differentiated from plants derived from self-pollination of Fleur 11 using morphological traits (dark green leaves and procumbent growth habit). The F 2 progeny were produced from the self-pollination of 15 F 1 plants. All crosses were performed in plastic pots under greenhouse conditions at the Centre d'Etudes Régional pour l'Amélioration de l'Adaptation à la Sécheresse (CERAAS) in Senegal during 2011 and 2013.

SSR Marker Analysis
Genomic DNA of both parents and F 2 progeny was extracted from young leaves according to the MATAB protocol as described by Foncéka et al. (2009). Polymorphisms were assessed in the parents using 602 primer pairs, mainly selected from a previous study (Foncéka et al., 2009). The parents and F 2 progeny were genotyped with polymorphic SSR markers. PCR was carried out in 96-wellplates in a total volume of 10 µl consisting of 5 µl of 5 ng/µl of the DNA template and 5 µl of a mixture of 0.1 µM of each SSR primer, 0.2 mM of each dNTP, 1X PCR buffer, 2.5 mM MgCl2, 0.1U/µl of Taq polymerase and 0.1 mM of IR700 or IR800-labeled M13 primer (MWG Germany) for fluorescence detection of SSR amplicons. A forward primer pair was labeled with a fluorescent (LI-COR Biosciences). Reactions were performed in an Eppendorf Mastercycler epgradient thermocycler. The PCR products were separated by electrophoresis run at a constant 95 W for 1−2 h in a DNA Sequencer (LI-COR 4300 DNA Analyzer, Lincoln, NE, USA).
Scoring of the SSR bands was performed visually on electrophoresis profiles using the application Jelly 2.017b (Rami, unpublished). For a codominant marker in an F 2 population, we denoted "A" as the genotype of cultivated parent, "B" as the genotype of wild parent, and "H" as the genotype of their heterozygous hybrids. When one marker was dominant, the two non-separated genotypes "H" and "A" were denoted "D" and the two non-separated genotypes "H" and "B" were denoted "C". For each SSR, the sub-genomic origin (A or K) of the wild alleles was assigned by comparing them to the alleles of the wild diploid progenitors. The sub-genomic origin of the cultivated allele was inferred by analyzing its co-inheritance with the wild alleles in the F 2 progeny. Loci were suffixed by <A> or <B> for differentiating subgenomic origin. In all scenarios, missing data were scored as "x" and the unexpected null bands were scored as "N".

Meiotic Behavior Analysis
Based on the genotyping data for each SSR marker, we analyzed the genotype of the F 2 progeny and deduced the allelic constitution of the gamete produced by the F 1 hybrid. The genome of the cultivated species was noted "A 1 A 1 BB" and that of the wild species "A 2 A 2 KK". Chromosomes "A 1 " and "A 2 " and "B" and "K" are homologous, while chromosomes "A 1 " and "B", "A 1 " and "K", "A 2 " and "B", and "A 2 " and "K" are homeologous.
When recombination occurred only between homologous genomes, for SSRs that amplified only one locus on a given sub-genome, two segregating bands were expected and the segregation ratio in the F 2 progeny was 1A 1 A 1 : 2A 1 A 2 : 1A 2 A 2 or 1BB: 2BK: 1KK. For SSR that amplified the two homeologous genomes up to four segregating bands were expected. The segregation ratio in the F 2 progeny is shown in case, the cultivated and wild parents produced each one type of gamete (A 1 B and A 2 K, respectively) and the F 1 hybrid (A 1 A 2 BK) would produce 4 types of gametes (A 1 B, A 1 K, A 2 B and A 2 K). When recombination occurred between homeologous genomes, for SSRs that amplified only one locus on a given sub-genome, a fourth 'null' genotype is expected to appear in addition to the three expected genotypes in an F 2 population. For SSRs that amplified the two loci, one on each of the homeologous genomes, 11 phenotypic bands are expected in the gel ( Table 1). The segregation ratio in the F 2 progeny depends on the genomic constitution of the F 1 hybrids. For F 1 with normal genomic constitution (without homeologous recombination during parental meiosis), the expected segregation ratio in the F 2 progeny is shown in Table 1. It is not straightforward to estimate the genotypic frequencies in F 2 progeny derived from "aberrant" F 1 (resulting from homeologous recombination events during parental meiosis). In contrast to the homologous pairings, two specific phenotypes, A 1 A 2 and BK, are observed in homeologous pairings. In this study, we use the terminology "homeologousrecombinant genotypes" for the progeny that carried these two peculiar phenotypes.

Allele Number in the Homeologous Recombinant Genotypes
In the electrophoresis profiles, the band intensities of the homeologous-recombinant progeny were analyzed to determine the number of copies of each allele. The band intensities were determined using ImageJ version 1.46 (Ferreira and Rasband, 2012), and the dosage ratio between bands was compared to the relationships expected between alleles in the hypothetical configurations. For example, for the homeologousrecombinant progeny with A 1 A 2 band-phenotype on the gels, three genotypes are possible with ratios of allelic combinations of 3:1 (A 1 A 1 A 1 A 2 ), 1:1 (A 1 A 1 A 2 A 2 ) or 1:3 (A 1 A 2 A 2 A 2 ). The genotype was then determined based on allelic dosage. The F 2 progeny with four different alleles (i.e., A 1 A 2 BK) were used as a reference for determination, as these progeny involve single-copy alleles.

Segregation Analysis
The segregation of codominant and dominant loci was compared to the segregation ratios expected under disomic inheritance (1:2:1 for codominant loci and 3:1 for dominant loci), using the chi-square test of the software "Calculation for the chi-square test" (Preacher, 2001). Data scored as "N" (null phenotype) as well as missing data were not considered in the analysis. Loci that deviated significantly (P < 0.05) from the theoretically expected ratios were considered distorted and were represented by an asterisk on the genetic map.

Linkage Analysis and Map Construction
Linkage analysis was performed using the MapDisto software (Lorieux, 2012) using only the co-dominant loci. The linkage map was constructed in several steps. In the first step, markers that showed distorted segregation (P < 0.05) and those with unexpected null data were excluded. The non-distorted loci were grouped into LGs using a minimum LOD of 3, a maximum recombination frequency r of 0.3 and the Kosambi mapping function (Kosambi, 1943). The order of markers within each linkage group (LG) was determined using the "order" and "ripple" commands. In a second step, the distorted loci and loci with unexpected null data were progressively added into the established LGs if their presence did not significantly affect the marker order. The position of the loci with unexpected null data was adjusted iteratively and the poorly mapped loci were removed. The synteny analysis between the homeologous genome was performed on the basis of the mapped homeologous loci (Foncéka et al., 2009). The graphical linkage maps were drawn using SpiderMap software (Rami, unpublished) and the graphical genotype was drawn using the GGT software (van Berloo, 2008).

Polymorphism of the SSR Markers and Segregation of Parental Alleles in the F 2 Progeny
Among the 602 SSR markers screened, 447 (74%) detected a polymorphism between the parental genomes. A total of 431 (71.6%) SSR loci were polymorphic between the A homologous genomes (A 1 and A 2 ) and 465 (77.2%) between the B/K homologous genomes. From these polymorphic SSR markers, 380 that provided accurate amplification profiles were analyzed in the F 2 progeny and generated 562 loci, out of which 378 (67%) segregated as codominant and 184 (33%) segregated as dominant. Among the 378 codominant loci, 194 (51%) were assigned to the A-genomes and 184 (49%) were assigned to the B/K-genomes. Among the 184 dominant loci, 70 (38%) were from the A-genomes, whereas 114 (62%) were from the B/K-genomes.
The segregation of the SSRs was assessed with regard to their informativity for distinguishing polysomic versus disomic inheritance. This allowed for the classification of the SSR into ten classes ( Table 2). The SSR markers within class 1 appeared as the most informative ones because they marked the homeologous genomes in both parents, allowing the identification of the expected genotypes for each of the meiotic behaviors in the F 2 progeny. They were chosen to illustrate the type of inheritance.

Inheritance Patterns of SSR Markers in the F 2 Progeny
The segregation analysis of 380 SSR markers revealed that 338 (88.9%) were inherited in an F 2 population, as expected under disomic behavior. Two compelling examples of disomic segregations are shown in Figure 1 for the SSR markers Seq8D09 and AC3C02. These primer pairs marked the homeologous genomes in both parents and amplified up to four segregating bands. Each band was assigned to a subgenomic allele ("A 1 ", "B", "A 2 " and "K"). The homologous bands of A-genomes ("A 1 " and "A 2 ") and B/K-genomes ("B" and "K") are inherited as two independent codominant loci in the F 2 progeny, Aa at one locus and Bb at the other (Figure 1). Genotypes were scored as "A", "B", and "H" for codominant loci and "B" and "D" (or "A" and "C") for dominant loci. a Indicates homologous alleles of identical size on the gel (monomorphism). b Dominant allele could be provided either by the cultivated or the wild parents.
FIGURE 1 | Images of the segregation of four SSR markers. The cultivated and wild parents are in lanes 1 and 2, respectively. The F 2 progeny are distributed from lane 3 to the end. Each primer pair amplified two segregating bands in both parents. Each band (allele) was assigned to a subgenome figured with arrows on the left. For the Seq8D09 and AC3C02 markers, the homologous bands inherited as two independent codominant loci in the F 2 progeny, i.e., A 1 A 2 genotype at one locus and BK at the other, as expected under disomic inheritance. At IPAHM108 and Ah3TC39B04 markers, the red and green arrows show the homeologous-recombinant F 2 progeny that own the BK and Aa by A 1 A 2 respectively. The homologous bands of the A-genome ("A 1 " and "A 2 ") are lacking whereas the homologous bands of the B/K-genome are systematically observed in 4 and 6 progeny (red arrows). Reciprocally, the "B" and "K" bands are lacking in one F 2 progeny, whereas, the "A 1 " and "A 2 " bands are observed (green arrow).
Interestingly, 42 (11.1%) SSR markers exhibited several unexpected F 2 band-phenotypes, which are impossible to explain under disomic behavior. Null bands for one of the homologous genomes were observed in some progeny, although the primer pair amplified the two homeologous genomes. Two examples of these unexpected F 2 phenotypes are shown in Figure 1 for the SSR markers IPAHM108 and Ah3TC39B04. These markers amplified up to 4 segregating bands. However, the homologous bands of the A-genomes ("A1" and "A2") are lacking for some F 2 progeny, whereas the homologous bands of the B/Kgenomes ("B" and "K") are observed, excluding the missing data assumption. Those F 2 progeny are indicated with red arrows on Figure 1. Reciprocally, for the Ah3TC39B04 marker, the homologous bands of the B/K-genomes are missing, while the homologous bands of the A-genome are observed in one F 2 progeny (marked with a green arrow on Figure 1). These A 1 A 2and BK-phenotypes are unexpected under disomic behavior, but are expected under polysomic behavior ( Table 1). These phenotypes occurred if and only if the homologous alleles ended up in the same gamete during the parental and/or the F 1 meiosis.
Remarkably, the value of the intensity of the "B" and "K" bands varied significantly among the BK-phenotypes for the SSR markers that segregated in a polysomic-like pattern, suggesting differences in the number of alleles. For example, in some BK-phenotypes, the more intense bands ("K") represent two and three doses, while the less intense bands ("B") represent a single dose (Supplementary Tables S1−S10). The K/B intensity ratio ranged from 1 to 3, indicating that the genotype of the BK F 2 band-phenotypes was either BBKK (1:1) or BKKK (1:3). The BKKK genotype is unexpected in the F 2 progeny derived from the A 1 A 2 BK F 1 hybrid ( Table 1). The presence of BKKK genotypes suggests the occurrence of homeologous recombination during the meiosis of the parents, particularly in the synthetic allotetraploid.
Moreover, for three polysomic SSR such as RN13D04, only one allele was observed in one F 2 progeny whereas two alleles were present in both parents with up to four alleles segregating in the population (data not shown).

Segregation Distortion Analysis
Among the 378 codominant loci scored in the F 2 progeny, 336 showed disomic inheritance. Of these, 279 (83%) followed the 1:2:1 Mendelian segregation ratio expected in an F 2 population and the remaining 57 (16%) loci deviated significantly from it (P < 0.05). A total of 17 and 40 loci were distorted for the A and B/K-genomes, respectively (Table 3). Of the 17 distorted loci for the A-genome, 14 were skewed toward the cultivated genotypes, whereas one and two loci were distorted in favor of the wild and heterozygous genotypes, respectively. Inversely, among the 40 loci distorted for the B/K-genome, 26 and 14 were skewed toward the wild and heterozygous genotypes, respectively. These results indicated differences between genomes and genotypes for SD.
All the 42 codominant SSR markers that showed an unexpected genotype under disomic inheritance were distorted when tested under the 1:2:1 Mendelian segregation ratio. We were not able to test for the segregation ratio expected under polysomic behavior (1:34:1) since the genomic constitution of the F 1 hybrids was unknown and unexpected genotypes could result from "aberrant" F 1 hybrids. Nonetheless, as indicated on the map below, these codominant distorted markers were clustered in particular regions on different LG, suggesting that such distortion likely originated from homeologous pairing.

Map Construction and Linkage Analysis
In our study, the genetic map was developed progressively due to the mixture in the dataset of codominant loci that segregated in a disomic or polysomic form, combined in some cases with the distortion of segregation. The map constructed without the distorted and the polysomic loci comprised 270 loci distributed on 16 LGs, while 34 loci remained unlinked. Distorted loci were then included in the map using LOD score values comprised between 4 and 7, r = 0.3. This allowed for the addition of two new LGs (LGB2 and LGB9) that were mainly formed by distorted loci (Figure 2). At this step, the map comprised 304 loci clustered in 18 LGs. In the third step, when adding the polysomic loci, two new LGs (LGA4 and LGB4) were formed and two initially distinct LGs (LGA3 and LGB3) clustered together. The clustered LGs were dissociated by removing three polysomic loci (Seq16F01, Ah3TC31H02, and Seq2H11). At this step, the resulted genetic map comprised 20 LGs, and 12 loci remained unlinked. The loci were then ordered within each LG. Finally, the estimated linkage map included 357 codominant loci distributed into 20 LGs and covering a total genetic distance of 1,728 cM, with an average interval of 5.1 cM between two adjacent markers.
We analyzed the distribution of the loci in the A and B/Kgenomes. For the A-genome, 179 loci were mapped in 10 LGs with an average number of 18 markers per LG, ranging from 10 (LG A5) to 36 (LG A1). The length of the LGs ranged from 28.0 cM (LG A5) to 165.2 cM (LG A1), with an average of 104.3 cM. For the B/K-genome, 178 loci were mapped on 10 LGs with an average number of 18 markers per LGs ranging from 13 (LG B10) to 28 (LG B1). The length of the LGs ranged from 33.3 cM (LG B10) to 111.4 cM (LG B1), with an average of 68.5 cM ( Table 4). The average interval between adjacent markers was 6.4 and 4.0 cM for the A-and B/K-genomes, respectively.
Except for three loci, all distorted loci (P < 0.05) were clustered along the LGs. Only 17 distorted loci were mapped in the A-genome, whereas 40 were mapped for the B genome. The distorted loci mapped in the A-genome were skewed toward the cultivated genotype (A4, A7, and A9 LGs), whereas those distorted in the B-genome were mainly skewed toward the wild genotype (B2 and B9 LGs) or the heterozygous genotype (B3, B4, B6, B7, and B10 LGs) (Figure 2). Overall, with few exceptions, we found that SD occurred in the regions that displayed homeologous recombination uncovering that homeologous pairing plays an important role in shaping SD in the tetraploid peanut genome.

Distribution of Disomic and Polysomic Loci along the LGs
One of the most remarkable features of this map is the distribution of the disomic and polysomic loci along the LGs. The green and red segments indicate the LGs deriving from the A genome, named from A1 to A10 and the B genome named from B1 to B10, respectively. The map distances in Kosambi map units (cM) of each LG are shown on the left, and the loci names are shown on the right. The LGs were grouped into homeologous pairs based on common homeologous loci connected by blue dashed lines. The polysomic and disomic markers are indicated in red and black colors, respectively. Duplicated loci are identified by the number 1 or 2 after the suffix A or B. The distorted loci are identified by asterisks after the locus name. The number of asterisks indicates the intensity of the distortion of segregation ( * P < 0.05, * * P < 0.01, and * * * P < 0.001). The symbols after the asterisk specify the direction of the distortion of segregation < >, < >, and < > indicate the distortion in favor of the cultivated genotype, heterozygous, and wild, respectively.
Among the ten homeologous sets of LGs, six sets consisted of markers that segregated in a likely disomic pattern and three in a mixed pattern with the disomic and polysomic loci clustered on the same LG (Figure 2). Surprisingly, one set (LG4) consisted of markers that all segregated in a polysomic way. The percentage of polysomic loci along the LGs ranged from 13.3% (LG6) to 100% (LG4) with an average of 39.9%.

Homeologous Recombination along the LGs and Homeologous Genome Substitution
The co-inheritance of the polysomic loci allowed for the estimation of the portion of the genome that underwent homeologous recombination. These portions ranged from 14.9 cM for LG6 to 95.5 cM for LG4, with an average of 43.7 cM ( Table 3). Our study clearly showed that for some progeny, the chromosomes were a mosaic of homologous and homeologous regions (Figure 3). Taking the physical distance covered by the LGs into account, a completed substitution of the A chromosome by its homeologous counterparts was observed in three F 2 genotypes (Figure 3).
The location of the homeologous recombination breakpoints along the chromosomes was assessed for LG4, which displayed a full homeologous pairing. Of the 17 regions in which the homeologous exchanges occurred, 10 regions of the B-genome were replaced by the A corresponding region and the seven others exhibited the reciprocal situation (Figure 3). The number of the homeologous recombination breakpoints along the LGs 4A and 4B ranged from 1 to 3.

DISCUSSION
The segregation patterns of SSR loci in tetraploid peanut (2n = 4x = 40) were determined by studying their inheritance and by mapping genetic exchanges between homeologous genomes. Our results strongly support a mixed disomic and polysomic modes of genetic inheritance of SSR loci in the cross between a cultivated peanut variety and a synthetic allotetraploid. Mixed inheritance is rarely observed in tetraploid peanut and has not yet been reported when the AAKK and AABB genomes were involved.

Evidence of Disomic and Polysomic SSR Inheritance
In our study, we showed that a large number of SSR markers were inherited likely in a disomic way, but some others showed genetic inheritance as a polysomic mode leading to a complex meiotic behavior.
In our study, to partially overcome the problem of complex electrophoresis profiles, we undertook a direct interpretation of SSR bands, suitably assigned to the subgenome of both parents. Using this approach, we observed that some markers followed the disomic inheritance as usually reported in allotetraploid peanut (Burow et al., 2001;Foncéka et al., 2009;Shirasawa et al., 2013) but some others exhibited an F 2 genotype unexpected under disomic inheritance, but that fitted with a polysomic segregation. The exclusive presence of homologous alleles among some F 2 genotypes for SSRs that mark homeologous genomes indicates that the homologous alleles ended up in the same gamete during the parental and/or the F 1 meiosis. Our findings are consistent with the recent study published by Leal-Bertioli et al. (2015) that explained the missing data observed among genotypes by tetrasomic recombination.
In some cases, the segregation patterns of SSR loci were similar to that observed in case of double reduction. We observed a FIGURE 3 | Graphical genotype of the 90 F 2 progeny at LG A4 and B4. Loci name is shown on the left. Each column represents a F 2 progeny. The green and red colors indicate the A and B/K genomes, respectively. The light-green, green, and dark-green colors indicate of the cultivated, heterozygous, and wild genotypes for A genome, respectively. The pink, red, and dark red colors indicate the cultivated, heterozygous and wild genotypes for the B/K genomes, respectively. phenotypic class with only one allele in some F 2 progeny, whereas up to 4 alleles segregate in the population. Double reduction refers to the fact that for a specific locus, the sister alleles come together in the same gamete during meiosis. It has been reported that the rate of double reduction is expected to increase towards the telomeres (Koning-Boucoiran et al., 2012;Bourke et al., 2015). In our study, the loci that showed this peculiar pattern of segregation are located at the extremity of LG3 (IPAHM103 and IPAHM282) and LG7 (RN13D04). However, since the genotype of the F 1 plant that gave rise to the F 2 progeny is unknown one cannot exclude that this pattern of segregation arose from the fusion of male and female gametes with the same haplotype (i.e., KK × KK).

Mapping Genetic Exchanges between Homeologous Genomes
The construction of the linkage map allowed us to locate the regions where the genetic exchanges occurred between the homeologous genomes. The LGs 3, 4, 6, and 7, which underwent homeologous recombination in our study, are consistent with the findings of Leal- Bertioli et al. (2015). However, the percentage of markers involved in homeologous recombination was higher in our study compared to that mentioned in Leal- Bertioli et al. (2015) (11% vs. 3%). Moreover, in LG4, we report a complete substitution of the A-chromosome by its B/K homeologous genome in three progeny. In the study of Leal-Bertioli et al. (2015), the LG4 of the induced tetraploid parent was involved in an almost complete substitution of the B-chromosome by the A-chromosome. The similarity of the results between these two studies using different synthetic allotetraploids raised questions about the factors that drive homeologous recombination. Pairing affinity between different sets of chromosomes was reported to be influenced by structural homology (Mason et al., 2010(Mason et al., , 2014Mandáková, Marhold and Lysak, 2014). Bertioli et al. (2016) reported a close similarity between the A and B wild species genomes based on sequence data comparison of A. duranensis and A. ipaensis. Moreover, in the present study and in many other genetic mapping studies in peanut (Burow et al., 2001;Foncéka et al., 2009;Shirasawa et al., 2013), a good collinearity was found between the homeologous genomes. These results are in favor of a homology-driven homeologous chromosome pairing.
However, we found some loci rearrangements between homeologous LGs that displayed a mixed inheritance. This was particularly the case for LG6, in which Bertioli et al. (2016) also reported a large inversion. Thus, there are probably others factors that drive the pairing between homeologous genomes in polyploid species. Those factors can be genetic (Cifuentes et al., 2010;Gaeta and Chris Pires, 2010), similar to that exerted by the Ph-1 locus in wheat (Moore, 2014) in rye (Lukaszewski and Kopecký, 2010) and Brassica (Nicolas et al., 2009). More studies will be needed to decipher the molecular forces that drive homeologous pairing in peanut.
The results of this study suggest that, at least in an interspecific context, the meiotic pairing in tetraploid peanut fits with the intermediate inheritance where pairing of chromosome sets ranged from strict disomic inheritance as in diploids to full polysomic as in autopolyploids (Stebbins, 1947;Sybenga, 1996;Catalán et al., 2006;Xu et al., 2015). Many important plants are polyploids and some of them have been clearly identified as intermediate genetic pattern such as chrysanthemum (Klie et al., 2014), banana (Jeridi et al., 2012), and strawberry (Lerceteau-Köhler et al., 2003). These findings also raised the question of chromosome pairing during meiosis in crosses involving only cultivated peanut varieties. Until now, inter-genomic exchanges have never been reported using genetic mapping approaches in cultivated peanut. Recently, thanks to tetraploid versus diploid sequence comparisons, genomic exchanges between A and B subgenomes have been reported (Bertioli et al., 2016). However, one still has to puzzle out whether these exchanges resulted from ancient recombination events that occured at the infancy of cultivated peanut as a consequence of the genomic shock due to the first hybridization between the A and B subgenomes or were due to something more recent that is still occurring in peanut varieties.

Homeologous Recombination Occurred during the Meiosis of the Parents
The reconstruction of the genotype of the homeologous recombinants suggests that, homeologous pairings have arisen in the synthetic allotetraploid parent. Indeed, the genotype of some homeologous recombinant F 2 progeny was either BKKK or BBKK. If homeologous recombination occurred only during the meiosis of the F 1 plants and not in the meiosis of the parents, the genotype of the homeologous recombinant in the F 2 progeny would only be A 1 A 1 A 2 A 2 or BBKK ( Table 1). The BKKK genotypic composition was regarded as a consequence of homeologous recombination during the meiosis of the parents particularly in the synthetic allotetraploid. In some species, homeologous pairing during meiosis is suspected to occur in the first generations following polyploid formation (Ramsey and Schemske, 2002;Soltis et al., 2010;Szadkowski et al., 2010).

Implications for Peanut Breeding
The pattern of recombination between chromosomes of related species is a key point to transfer genes between the species. The knowledge about the inheritance mode is essential information not only because it sheds light on homeologous chromosome pairing behavior but can also influence the breeding strategies that are used for cultivar development. In this study, homeologous recombination events have been located in some genomic regions that are common with the ones reported in the study of Leal-Bertioli et al. (2015). One can suppose that these regions are particularly sensitive to homeologous recombination especially when synthetic allotetraploid are used. The SSRs that are located in these specific regions, particularly those that marked the homeologous genome in both parents, have proven to be very efficient in revealing the homeologous recombination events. Thus, they can be used to detect and trace back those events in the synthetic allotetraploid parental lines before crossing, as well as in their related F 1 hybrids.
Mixed inheritance may have unwanted impacts on aberrant meiotic behavior, karyotype destabilization, and fertility reduction. However, it could also speed up the accumulation of rare but favorable alleles through homeologous recombination and marker-assisted introgression. The interspecific breeding population developed in this study would be an ideal genetic material to study the genetic effects of cumulative homologous and homeologous alleles and to transfer valuable alleles from wild species into cultigens. Despite conflicting results around the A. batizocoi taxon and its potential usefulness , its utilization for peanut breeding has been successful reported (Burow et al., 2014;Sukruth et al., 2015). Although erratic fertility was observed in some lines, the advanced backcross population developed from the same cross has a high phenotypic variation for many important agronomic traits, such as plant architecture, yield related traits, drought tolerance and resistance to leaf spot (Nguepjop et al., in preparation). We believe that the genetic variation among AB-QTL lines is increased when homeologous chromosomes pair. The mosaic compositions of the genome and the homeologous chromosome substitutions may speed up novel genetic combinations, opening new horizons for peanut breeding.

CONCLUSION
The inheritance patterns of SSR markers, statistical analysis and genetic mapping provide evidence of a mixed disomic and polysomic mode of genetic inheritance in allotetraploid peanut based on an experimental interspecific cross. The mixed inheritance appears associated with segregation distortion and homeologous chromosome substitutions. These findings contribute to a better understanding of the meiotic behavior of allotetraploid peanut and will provide useful information to breeders that use synthetic tetraploid to move genes in the genetic background of the cultivated peanut species.

AUTHOR CONTRIBUTIONS
JN designed and coordinated the study, performed the experiments, carried out data analyses and map construction, and wrote the manuscript. H-AT was involved in population development. NM and SS have produced the synthetic amphidiploid used in the study. JB, BC, and DS were involved in the design of the study and helped in data analysis. J-FR designed the study involved in map construction and contributed to editing of the manuscript. DF conceived, designed, and coordinated the study, helped in data analysis and editing of the manuscript. All authors read and approved the manuscript.

FUNDING
This work was funded by WAAPP/PPAAO Senegal.