QTL Mapping for Resistance to Cankers Induced by Pseudomonas syringae pv. actinidiae (Psa) in a Tetraploid Actinidia chinensis Kiwifruit Population

Polyploidy is a key driver of significant evolutionary changes in plant species. The genus Actinidia (kiwifruit) exhibits multiple ploidy levels, which contribute to novel fruit traits, high yields and resistance to the canker-causing dieback disease incited by Pseudomonas syringae pv. actinidiae (Psa) biovar 3. However, the genetic mechanism for resistance to Psa observed in polyploid kiwifruit is not yet known. In this study we performed detailed genetic analysis of a tetraploid Actinidia chinensis var. chinensis population derived from a cross between a female parent that exhibits weak tolerance to Psa and a highly Psa-resistant male parent. We used the capture-sequencing approach across the whole kiwifruit genome and generated the first ultra-dense maps in a tetraploid kiwifruit population. We located quantitative trait loci (QTLs) for Psa resistance on these maps. Our approach to QTL mapping is based on the use of identity-by-descent trait mapping, which allowed us to relate the contribution of specific alleles from their respective homologues in the male and female parent, to the control of Psa resistance in the progeny. We identified genes in the diploid reference genome whose function is suggested to be involved in plant defense, which underly the QTLs, including receptor-like kinases. Our study is the first to cast light on the genetics of a polyploid kiwifruit and suggest a plausible mechanism for Psa resistance in this species.


Introduction
Most angiosperm species are diploid, the state where there are two complete sets of chromosomes in each nucleus. However, many plant species are polyploid, exhibiting more than two complete sets of chromosomes [1][2][3][4]. A number of agricultural and horticultural crops have their origin in polyploidy or are polyploid [5,6], for example banana, coffee, potato, cotton, strawberry and wheat. However, polyploids present a significant challenge to geneticists for the investigation of the genetic basis of traits, owing to the complex segregation patterns and preferential vs non-preferential pairing among chromosomes observed in such species [7][8][9]. Recent progress in sequencing and genotyping methodologies, as well as statistical tools for analyzing genetic data from polyploids, have overcome key issues that previously hindered the understanding of polyploid genomes [10,11].
The manner in which different homologous chromosomes pair and recombine during meiosis further determines the biological properties of a polyploid genome and the transmission of genetic information to the offspring. For example, in a tetraploid species where there is random pairing of homologous chromosomes, recombination occurs among all homologues. This manner of inheritance is termed polysomic and is characteristic of auto-polyploids such as potato and alfalfa, but can also be observed in "segmental" allopolyploids, such as peanuts [12][13][14]. Where the four homologous copies of a chromosome in a tetraploid species pair and recombine strictly preferentially, this results in disomic inheritance [13,14]. This pattern is typical of allopolyploids, such as wheat and coffee. In general, chromosomal pairing preference provides strong clues as to the origin of chromosomes in a species and its progenitors.
Pure auto-and allopolyploids are the two extremes of the spectrum and there are several species that fall in between, including some natural polyploid populations in the genus Actinidia (kiwifruit), which appears to be a "segmental" allopolyploid [15][16][17]. Additional polyploid kiwifruit selections have been generated by Plant & Food Research's breeding programme by artificial chromosome doubling and crossing between different ploidy levels. Segmental allopolyploids can originate from parents belonging to distinct subspecies [18], resulting in the formation of templates for divergent genomes, with repercussions for chromosomal as well as transcriptional and epigenetic modifications [19][20][21]. This form of hybridity in genomes can contribute to homologous recombination among segments of chromosomes from respective parents [22], which entails changes at the molecular and phenotypic levels, paving the way for adaptation, speciation and invasiveness. In New Zealand, some of the native flora as well as introduced agricultural plant species exhibit allopolyploidy [23].
The exotic kiwifruit, with their high vitamin and mineral contents, attract consumers and contribute billions of dollars to the global economy, in addition to supporting the New Zealand horticultural industry. Originating in China, the Actinidia genus holds a resource of ploidy levels recognized at the sub-species level [24]. The green-fleshed hexaploid (2n = 6x = 174) A. chinensis var. deliciosa ('Hayward') was first cultivated in the mid-19th century, followed by the yellow-fleshed diploid (2n = 2x = 58) A. chinensis var. chinensis ('Hort16A') in the 1990s. 'Hort16A' succumbed to the global pandemic incited by the canker-causing Pseudomonas syringae pv. actinidiae (Psa) biovar 3 [25,26]. The genomes of diploid A. chinensis var. chinensis [27][28][29] and A. eriantha [30] have recently been sequenced, which will significantly assist genetic and genomic studies associated with agronomic traits, as well as elucidating the evolution of the genus and its fitness over diverse climates and geographies.
Breeding programmes focus on tetraploid cultivars, because of their resilience to the Psa disease, as well as robust fruit quality traits and size. However, there remains a substantial knowledge gap concerning the chromosomal biology and complex gene-trait associations in polyploid Actinidia compared with diploid species. Actinidia species are mostly dioecious and exhibit high heterozygosity [24]. This high degree of heterozygosity can facilitate trait mapping in segregating F1 populations using quantitative trait locus (QTL) mapping or association studies [11]. New genotyping platforms, including genotyping-by-sequencing (GBS) and single nucleotide polymorphism (SNP) arrays, are now available to capture allelic variations and haplotypes across the genome [31]. Furthermore, new tools have emerged for utilizing the abundant sequence data generated for the purpose of dissecting the genetic determinism of complex traits in polyploids, including the software TetraploidSNPMap [32] as well as packages based in R that include polymapR [33], net was [34], MapPoly [35], QTLpoly [36] and PERGOLA [37].
In this study, our goal was to cast light on the genetic regulation of resistance against the canker-causing Psa biovar 3 in a tetraploid F1 population of kiwifruit, by performing QTL mapping based on disease phenotype data collected in the field and greenhouse. A previous study has suggested that Psa resistance in a diploid yellow-fleshed kiwifruit population is governed by several QTLs [38]. However, the greater resilience against the pathogen, as well as low severity of disease symptoms compared with those in diploids [39], raises questions about the mode of inheritance of resistance in tetraploids. In the present study we utilized a robust new approach based on capture-based sequencing (Capture-Seq) to generate high density linkage maps using polymapR [33], followed by polyqtlR [40] for performing QTL mapping using identity-by-descent trait mapping [40]. Our analysis provides a first and comparative view of the genetic landscape of a tetraploid A. chinensis, and insights into the genomic regions regulating the quantitative form of Psa resistance in a yellow-fleshed sub-species of kiwifruit.

Bait Selection for High-Density SNP-Based Genotyping in Kiwifruit
A total of 9918 baits were designed by Rapid Genomics (Gainesville, Florida, USA) across the 29 chromosomes of diploid kiwifruit Red5, to genotype the population using Capture-Seq. These are exome baits designed within the genes to capture regions of 120 bp in length. Within each chromosome, 10-20 baits were spaced on average within every 1 Mb region. On average there were~333 baits per chromosome ( Figure S1).

Variant Calling and Dosage Estimation
A total of 9918 baits were used to genotype 235 F1 individuals of the tetraploid A. chinensis P1 × P2 population. These generated 1.03 × 10 9 paired-end reads with a high quality score (mean Phred score 37). The population had a mean and median depth of 120x and 111x across baits, respectively. Within this dataset, a total of 725,175 raw SNP variants were identified that had an alignment rate of~94.5% to the Red5 diploid kiwifruit genome. Estimates for self-relatedness in the population, calculated using the SNP dataset and read depth adjustments, indicated 20 individuals were outliers, suggesting that pollen contamination had occurred ( Figure S2). After the removal of these contaminants, 13 F1 individuals with low quality (lower sample depth and call rate) were removed ( Figure S2), which resulted in a final set of 188 true hybrids with high-quality sequence data for variant selection (Figure 1a).
A plot of the mean sequencing depth against the proportion of individuals with at least one sequence read at each SNP position (SNP call rate) indicates that most genotypes were called with sufficient read depth (1.00 log scale) (Figure 1b). A substantial proportion of SNPs had high minor allele frequency (MAF) (0.5), although SNPs with a lower MAF (0.275) were also common ( Figure S3). Characterization of allele types by an allele frequency count in one of the seedlings demonstrated allele frequency rates in the dataset and was employed to represent the ploidy level of genotypes. Figure 1c,d highlight the frequency peaks of homozygous alleles/haplotypes (0/0 and 1/1) and heterozygous alleles/haplotypes (1/2, 3/4 and 1/4), respectively, for tetraploids. The 1/3 fractions in allelic frequencies in tetraploids have been previously noted and could point to possible null alleles (aaA0 or aAA0) [41] or other phenomena such as copy number variations in the repetitive portion of the reference genome [42]. Filtering of SNPs to a minimum read depth of 60 yielded~70,000 markers usable for dosage estimation. (b) is a plot of SNP call rates (which is the proportion of individuals observed with at least one allele) versus mean SNP depth (log scale); (c) shows allele frequency in one of the F1 genotypes for the most common class of homozygous variants (0 and 1); (d) shows heterozygous variants in the same genotype as in (c), with peaks at 1/3 and 1/2. Allele frequencies are 0, 1/4, 1/3, 1/2, 1/3, 3/4 and 1; (e) shows three different plots for genotype dosage estimates in F1 individuals for three different bi-allelic SNPs (1-3). The dots represent individuals between the read counts for the reference allele (A, y-axis) and the counts of reads for the alternative allele (a, x-axis). The dots are colored based on the five genotype dosages in tetraploids (ploidy+1) such that the genotypes 0, 1, 2, 3 and 4 refer to dosages aaaa, Aaaa, AAaa, AAAa and AAAA respectively. An empirical Bayesian analysis (Updog, "normal model") to estimate genotype dosage in F1 individuals assists the visualization of the allelic dosage for SNP haplotypes. For bi-allelic SNPs, there are 5 possible dosage classes (ploidy +1) which are 0 (aaaa), 1(Aaaa), 2(AAaa), 3(AAAa) and 4(AAAA), where 'a' refers to the alternate allele while 'A' corresponds to the reference allele ( Figure S4). Figure 1e comprises of 3 plots from Updog output, each demonstrating which genotype dosage is carried by the F1 individuals for a particular SNP. For example, each of plot 1 and 2 in Figure 1e demonstrates that the 188 F1 individuals are called to discrete classes as uniformly as possible for the respective SNPs, i.e., AAAA and AAAa in plot 1 and AAAA, AAAa and AAaa in plot 2. However, in plot 3 in Figure 1e we can observe that the dosage for the two major clusters of individuals is AAAA and AAAa, while a few individuals are called for genotype dosage 2, i.e., AAaa. This was not expected and could probably be due to systematic biases or over-dispersion for this particular SNP and hence qualifies for removal. Subsequent filtering of SNPs using Updog for a range of quality metrics, such as allele bias, over-dispersion and sequencing error rate, resulted in 60,006 usable bi-allelic markers representing 24 different marker types (Figure 2a).
The sex locus for kiwifruit is located on chromosome 25 [43,44], making this one of the chromosomes of particular interest ( Figure 2c). We calculated the goodness-of-fit measure of the linkage maps by comparing multi-point estimates of pairwise recombination frequency from the genetic map, with the two-point estimates of recombination frequency. The weighted root mean square of the difference between these two estimates showed an inverse relationship with the logarithm of the odds (LOD) scores ( Figure 2d). Where the difference between the expected and estimated recombination frequency was high, the LOD values were low (Figure 2d), indicating that the overall map quality was good. Figure 2e shows a plot of the position of markers on LG25 versus the estimate of recombination frequency with all other markers, demonstrating that tightly linked markers exhibit lower recombination frequency at the diagonal, depicted in green. Hence the estimated LOD score from the adjacent markers on the map would also be high, as represented by the red color in Figure 2f.
All 29 framework LG maps are depicted in Figure 2g, where each of the maps represents the integrated map derived from the phased maps of eight homologues, four from each parent. A total of 39,322 markers were assigned across the LGs, making this one of the highest-density linkage maps published for a polyploid species. The map lengths ranged from minimum 91 cM to maximum 143 cM, with a mean length of 115.5 cM. Genetic maps of LGs between 125 and 250 cM have been observed in various diploid, tetraploid and hexaploid species [45,46], with higher values often indicative of poor data quality or high map stress. Overall, the LG map lengths from this study indicate that a high-quality genetic map has been constructed. To further assess the quality of our maps, we estimated the Identity-by-descent (IBD) probabilities [47,48] for nine seedlings in the population ( Figure 3). Predicted allelotypes shown in dark colors have high probability of inheritance, while longer segments of parental haplotypes with relatively few light-colored bars represent regions of recombination breakpoints, which when taken together indicate a high quality dataset for QTL mapping.

Comparison of the Physical and Genetic Maps of Tetraploid A. chinensis and Quantification of Preferential Pairing
Marey plots enabled us to identify the degree of synteny between the tetraploid and diploid A. chinensis Red5 genome at the LG level ( Figure 4). The baits used to genotype the tetraploid A. chinensis population carried positional information from the physical chromosomal scale assembly of the diploid kiwifruit genome Red5 [28], which was compared with the position of derived markers on the genetic maps.
We found that in most LGs, the marker positions as derived from the physical map of the diploid genome exhibited linear correspondence with the genetic positions in the tetraploid map. This indicates that, since polyploidization in A. chinensis, there have been only a few structural changes between the diploid and tetraploid genomes and that the order of the genes on the chromosomes of tetraploid A. chinensis may not be very different from that of the diploid A. chinensis.
However, the Marey plots suggest that for some chromosomes this situation does not apply for small segments of sequence, i.e., the marker genetic position does not follow the linear order of the physical map of Red5. These disruptions were visible on the plots for LG7, LG8, LG10, LG17, LG19, LG23 and LG27 and could arise from reads aligning to duplicated regions, from translocation events in the tetraploid genome, or possibly from artefacts from the mapping procedures (either in the physical assembly or the genetic map, or both). The Marey plots also depict regions on maps where the linear correspondence between the genetic and physical position of markers is weak, suggesting repressed recombination rates, for example the upper arm of LG25 (Figure 4), which corresponds to the sex-determining region and has been shown to have suppressed recombination rates in diploid A. chinensis [43,44]. More generally, these regions can be interpreted as putative sites of pericentromeric heterochromatin (i.e., the centromeres). Finally, we tested the nature of pairing among chromosomes in tetraploid A. chinensis using our marker dataset. Our results show that pairing was random across most linkage groups, except for a few LGs with behavior deviating towards preferential pairing (at X 2 < 0.001) ( Figure 5). This included LGs 20, 23 and 27 in P1 and LGs7, 15, 22 and 27 in P2.
LG27 exhibited preferential pairing in both parents, while LG3 in P1 has a strong but statistically weaker signature for preferential pairing. Hence, we find that the inheritance in tetraploid A. chinensis is mostly polysomic with preferential pairing in a few LGs. This is consistent with the cytological findings in tetraploid A. chinensis described earlier [16], and supports the view that tetraploid kiwifruit is a "segmental" allopolyploid.  Figures (a,b) show ranking of linkage groups (LGs) for pairing in Parent 1 and 2, respectively. Pairing is calculated by X 2 values generated by polyqtlR for each LG in each parent, with an arbitrary threshold of X 2 < 0.001 set as indicative of preferential pairing. This is represented by the yellow dotted line and LGs above it are considered to show polysomic inheritance whereas LGs below it are likely to be disomic. The log 2 (X 2 ) is plotted on the Y-axis with LGs on the x-axis differentiated by color. The size of points in the scatter plot is determined by the number of individuals from the population that were used for each parent and referred to as F1-P1 or F1_P2. These individuals showed an unambiguous pairing configuration.

Field
The population was scored for resistance to Psa in two different orchards over the two years after planting. Best linear unbiased predictor (BLUP) values for Fld_Psa_scores were not Normally distributed (Figure 6a), with skewing towards the right of the order of −0.15 showing towards more disease. A non-Normal distribution was previously reported for the Fld_Psa_score in a diploid A. chinensis population [38]. Correlation of field observation breeding values between sites was high (r 2 = 0.61), with between-site clonal repeatability of 0.75 indicating a strong genetic effect and a smaller effect for site in the variance component analysis.

Stab Assay
Evaluation of five clonal replicates of each individual genotype in the population for Psa tolerance under greenhouse conditions inoculated using the stab assay revealed four key phenotypes characterizing the response to Psa infection. These included Psa-induced stem necrosis, ooze, wilt, and leaf spot, as well as a combined Psa score for each seedling in the population, calculations based on all four phenotypes. The distribution of Stab_Psa_score and Stem_necrosis across the population was Normal, whilst data for ooze, wilt and leaf spot were distributed in a non-Normal manner (Figure 6b-f). Normal distribution of the Stab_Psa_score was previously reported in a diploid A. chinensis population [38]. The analysis showed that greenhouse inoculation results are highly dependent on time of inoculation, similar to the earlier results for the diploid population [38], with clonal repeatabilities substantially lower than field repeatabilities. The best inoculated repeatability values were found for the sets inoculated in June (0.5) (winter) and the lowest for the sets inoculated in January (0.3) (summer).
Correlation between the greenhouse stab inoculation scores and combined field scores were poor (r 2 = 0.04), indicating that the variables and responses measured in the greenhouse and field were different and that the ranking for the greenhouse inoculation set was a poor predictor of the field observations. Broad sense heritability values for greenhouse disease ratings calculated from seedling means were lower (0.0 to 0.51) and more variable than for the field observations. This finding was probably due to the large random effect found for inoculation time (owing to space limitations, higher temperatures in summer [49] and developmental differences, not all seedlings were inoculated at the same time of the year).

Sex Locus
Because of their dioecy, the sex phenotype of kiwifruit plants plays an important role in commercial crop production. QTL mapping for the sex locus has not previously been performed in tetraploid kiwifruit and this provided us with the opportunity to validate tetraploid QTL mapping, using the flowering phenotype, before mapping for resistance. Flowering in kiwifruit over a whole population occurs in the third year of planting. However, 77 genotypes in the mapping population flowered in the second year, including 43 females and 28 males, enabling us to use this dataset to map the QTL associated with sex. We found a strong QTL on the upper arm of LG25 at 9.1 cM (LOD 17.2, PVE (percentage variance explained) = 64) over the marker Chr25:2290069 (Figure 7a). This region has previously been shown to contribute to the sex phenotype of the kiwifruit plants in diploid A. chinensis and the genes controlling sex-determination have been located on the Y-chromosome, LG25 [43,44]. The mapping of the phenotype to homologue 6, inherited from the male parent P2, can be clearly seen from the haplotype trace (as indicated by the high positive value associated with strong green color) in the IBD profile for LG25, below the QTL profile (Figure 7a).
The genotypic-information-coefficient (GIC) [47] was estimated across the population for each linkage group and used to assess the reliability of QTL estimation (Figure 7b). Where the GIC is close to one, estimates for unbiased QTL detection within that particular region are considered highly reliable. The GIC estimation for LG25 for both parents also demonstrates that GIC values can be variable at the telomeric region. This can affect the accurate positioning of a QTL in this area [47]; however, in the present case GIC is close to one at this position (~10 cM) on most homologues, including homologue 6. We can conclude that our trial QTL mapping for the sex phenotype in our tetraploid population identifies a male-specific genetic region on Y-chromosome that has been previously mapped in a diploid population and is used currently to predict sex for pre-breeding, clearly demonstrating that QTL mapping has been effective in our population. Our approach for studying the genetic basis of qualitative and quantitative traits in a tetraploid kiwifruit population was based on recent advances in QTL mapping in plant species that exhibit higher ploidy, and achieved using high-throughput genotyping with Capture-Seq, as well as access to new tools for genetic and QTL mapping for complex genomes of plant species [10,46,47,50,51].

Psa Resistance
QTL mapping of Fld_Psa_score, using data from two years of phenotyping, revealed four QTLs for resistance to Psa in the field. The first QTL is located on LG1 at 68.1 cM (LOD 4.5, PVE = 11.3), over the marker Chr1: 14996600 (Figure 8a). The 1-LOD support interval for this QTL lies between 58 and 72 cM. The second QTL is located on LG2, with its peak at 93.3 cM (LOD 5.0, PVE = 9.7) over marker Chr2:13522420 and a 1-LOD support interval between 89 and 99.1 cM (Figure 8b). The third QTL is located on LG4 at 54.4 cM (LOD 4.5, PVE = 11.2) over the marker Chr4:5917168, with the 1-LOD support interval for this QTL located between 45 and 63 cM (Figure 8c). The final QTL peak is detected on LG7, at 59.2 cM (LOD 5.1, PVE = 12.8) over underlying marker Chr7:6815185. The estimated 1-LOD interval support for this QTL peak is between 53 and 63 cM. These QTLs were significant at p < 0.05 and p < 0.1 using a chromosome-wide permutation test (Figure 8d). The contribution of homologues under each of the QTLs is indicative of the source of Psa resistance in the population. Taken together, P2, which is the more resistant parent, has three QTLs, including one at h8 under LG1, and two at each of the h5 of LG4 and LG7 which contributes towards field resistance or quantitative resistance (QR) against Psa, while P1, which is weakly tolerant, also contributes a minor QTL for Psa resistance on h2 of LG2. Previously it was shown that in diploid kiwifruit, a QTL for resistance to Psa is located in a Psa susceptible cultivar 'Hort16A' [38]. Similarly, other studies have pointed out towards alleles from susceptible parents involved in delivering varying degree of resistance against different pathogens [52][53][54].
To determine if the QTLs act in combination in delivering QR we next calculated the PVE combined, to assess the additive effect of all four QTLs in combination. The 4-QTL model (Q1(LG1)_Q2(LG2)_Q3(LG4)_Q4(LG7)) indicated a total PVE of 34. We suggest that the four QTLs act in combination in the progeny to provide a more robust QR against Psa in the field. Notably, these genetic regions do not overlap with those identified in previous work on locating QTLs controlling QR against Psa in the yellow-flesh diploid A. chinensis [38], suggesting that the origin of Psa resistance in tetraploid species is not related to any common ancestral loci in Actinidia species and that the sources of QR might be diverse in terms of their positions in the kiwifruit genome in between the sub-species.
Our results for QTL mapping for control of resistance to Psa using data from the field suggest that Psa resistance in the tetraploid A. chinensis population is a polygenic trait which is likely to be controlled by multiple genes. This is consistent with a previous hypothesis for mode of Psa resistance in tetraploid kiwifruit populations [39]. We further demonstrated which key homologues from P2, as well as P1, contribute to these QTLs that act in combination to exhibit Psa resistance in the field. Further validation of the QTLs from the Field_Psa_score needs to be performed in backcrosses as well as out-crosses in tetraploid populations which share either P2 in their pedigree or grandparents in the F1 population.
QTL mapping of the response to Psa inoculation measured in the stab assay identified two major QTLs for the Stab_Psa_score on a pseudo-autosomal region on LG25 (LOD 6.9 at 107.1 cM with underlying marker Chr25:19357486) and LG29 (LOD 5.9 at 73.4 cM over marker Chr29:11615148) ( Figure S5). One QTL was identified for Stem_necrosis on LG21 (LOD 5.2, at 37.8 cM over marker Chr21:3833187), as well as a single QTL for Leafspot on LG16 (LOD 4.3 at 34.8 cM over marker Chr16:3574359) ( Figure S6). Notably, no significant QTLs were detected for the Ooze and Wilt phenotypes. The QTLs for phenotypic data from the Stab assay did not overlap with the QTLs from the Fld_Psa_Score. We found a similar scenario in our previous study, where QTLs from the stab assay did not overlap with the QTLs for Psa resistance in diploid A. chinensis [38]. It seems plausible that the expression of Psa resistance is influenced by the mode of infection in the stab assay, compared with the field phenotyping, as well as being intricately dependent upon developmental stage, environment, and tissues. Just below the QTL profile is information on the location on the genetic map of the alleles contributing to the phenotype, as well as the source of the alleles from the parental homologues. P1 homologues are represented as h1-h4 and P2 homologues as h5-h8. The positive score/effect of the alleles from the respective homologue is identified by the strong green color, and the quantitative measure for this effect is provided in the legend beside the QTL graph. In addition to QTL information, the graphs are supplemented with the GIC profile of each parent, in two separate panels for each LG.

Candidate Genes for Control of Field Psa Tolerance Underlying the QTLs
The genes determining the QTLs identified using the Fld_Psa_score phenotype are expected to play an important role in determining the genetic basis of control of Psa resistance in tetraploid A. chinensis. Our search among the gene models in the diploid Red5 genome, underlying the genetic markers flanking the four major QTLs yielded genes related to signaling, transporter, and various other functions related to carbohydrate metabolism, transcription and stress tolerance. Table S1 lists all the predicted gene models identified on the diploid Red5 genome at the locations underlying the four QTLs. From these we have highlighted a few, most of which are associated with plant immunity, including Acc00939.1 (Proline-rich receptor-like protein kinase PERK9-like), Acc00941.1 and Acc00956.1 (Receptor-like serine/threonine-protein kinase ALE2) as well as Acc00949.1 (Wall-associated receptor kinase-like) at LG1; and Acc04873.1 (WD and tetratricopeptide repeats-like), and Acc04875.1 (Serine/threonine-protein kinase-like) Acc04916.1 (Sugar transport protein 5) on LG4. Five genes are annotated on LG2 with putative functions associated with plant immunity to pathogens. These include Acc02420.1 (Glycerol-3-phosphate acyltransferase-like), Acc02421.1 (a Protease Do-like 1, chloroplastic-like), Acc02426.1 (Serine/threonine-protein kinase-like), Acc02438.1 (High affinity nitrate transporter-like) and Acc02443.1 (Mitogen-activated protein kinase 3-like). A further five genes involved in plant-pathogen interaction were identified under the QTL on LG7, including Acc07849.1 (Transmembrane protein adipocyte-associated 1-like), Acc07852.1, Acc07869.1 and Acc07861.1 (Serine/threonine-protein kinase EDR1-like) and Acc07858.1 (WD repeat-containing protein 48-like).
Overall, our results indicate that most genes underlying QTLs for control of resistance to Psa in tetraploid A. chinensis belong to a gene network that encodes proteins which have putative functions in membrane and intracellular signaling, such as receptor-like (RL) serine/threonine kinases and anion transporters, as well as putative enzyme functions. There seems to be functional commonality in the genes that underlie the QTLs identified for Psa resistance in the field in both diploid [38] and tetraploid A. chinensis; for example, the gene family that encodes RL kinases. RL kinases are known to play a significant role in plant immunity, especially early pathogen recognition and activation of plant defense [55][56][57][58]. The mechanistic basis of resistance for the two sub-species may rely on combinatorial action of defense proteins providing strong surveillance for pathogen-associated molecular patterns (PAMPs) and activation of signaling cascade for PAMPS-triggered immunity [59]. Our findings pinpoint gene families, genetic regions and mechanisms underlying resistance to Psa in diploid [38] and tetraploid yellow-fleshed Actinidia species.

Plant Material
The F1 population for genetic mapping of resistance to Psa was developed from a cross between a moderately Psa-tolerant female 'P1' and a highly resistant male 'P2'. The designations and a pedigree are provided for these parents in Figure S7. A total of 235 F1 genotypes were germinated aseptically in 2017, under standard tissue culture growth conditions. Each seedling was replicated 15 times by cuttings, either using tissue culture, or under standard greenhouse conditions, prior to phenotyping in the field, or with stab bioassays. The mapping population was maintained under standard orchard conditions, without any control for Psa infection, in both research orchards for two years. All the seedlings which were replicated for the field trial were also replicated five times for the stab assay. Plants were grown up to 1 m in height in pots in the greenhouse and then subjected to the stab assay in batches, as described previously [38,60].

Phenotyping
The field population was assessed for symptoms in response to natural Psa infection between 2018 and 2020. Traits scored included cane death, ooze, shoot death and tip death. A cumulative Psa score (Psa_score_Field) was calculated after two years, as described previously [38]. Least Square Means (LSMs) were calculated for this score for each genotype and BLUPs were estimated.
The stab bioassays were performed as described previously [38,60], in standard greenhouse conditions between September to December (spring) and February to April (summer) for two years under standard greenhouse facilities at Plant & Food Research Ruakura, Hamilton. Inoculations were performed with the Strain 10627of Psa biovar 3 [61], in the greenhouse at temperatures ranging between 22 and 30 • C. The flowering data for the population were obtained in 2019.

Genotyping, Genetic Map Construction and QTL Mapping
DNA was extracted from freeze-dried leaves using the Cetyl trimethylammonium bromide (CTAB) method [62] and~1 µg of DNA was provided to Rapid Genomics LLC, Gainesville, Florida for Capture-sequencing. The DNA library preparation and capture enrichment were done by Rapid Genomics LLC). They designed approximately 10,000 baits in the Red5 genome [28] to capture genetic variations, targeting putative SNPs, using high throughput Illumina ® sequencing, which generated 150-bp paired-end reads. Raw reads were assessed for quality using FASTQC (https://github.com/sandrews/FastQC) and trimmed using Trimmomatic [63]. Reads were aligned to the Red5 genome using Bwa-MEM [64] and Samtools [65].
Linkage mapping analysis and phase calculation were performed using polymapR [33] (1.1.0), which further filtered the dataset for high quality SNPs (qall_mult > 0 using function checkF1). Linkage maps were built chromosome by chromosome, owing to the high number of genetic markers and linkage groups and the amount of computing required. Once generated, the individual Linkage Group (LG) phased maps were merged for QTL mapping using polyqtlR [40]. This software performs QTL mapping in the F1 populations of outcrossing, heterozygous polyploid species and relies on identity-by-descent (IBD) probabilities to perform interval mapping, following the approach proposed by Kempthorne [70] and later developed by Hackett [71,72]. The software was made available to this project by Peter Bourke and will soon be available through the Comprehensive R Archive Network (CRAN: https://cran.r-project.org/). Preferential pairing among LGs was calculated using the IBD estimation function polyOrigin, which is part of the polyqtlR package, using the default function test_prefpairing. Percentage variance explained by each QTL and the combined model was calculated using the PVE function in polyqtlR. Owing to the large number of LGs and high density of markers, LOD thresholds were calculated for each LG separately, using a permutation test (1000 permutations, at α <0.01 and <0.05). Genes underlying the QTLs were identified from the in-house kiwifruit gene model database for the diploid Red5 genome version 1.69.0 [28]. The positions for the genomic regions were derived from the location information present within each marker.

Conclusions
Our study has demonstrated the application of capture-sequencing to generate a high density SNP-based linkage map in a heterozygous polyploid fruit crop. This is the first study to report the genetic maps of a polyploid Actinidia species. The data showed that the tetraploid kiwifruit A. chinensis exhibits polysomic inheritance for most LGs, as well as disomic inheritance for several LGs, fulfilling the characteristics of a 'segmental' allopolyploid. IBD-based QTL mapping for control of the response to Psa infection in two different field sites enabled us to map four QTLs for control of Psa resistance in the field in this population, identify the alleles on parental homologues and the additive effect exhibited in the variance. The candidate genes underlying the QTLs will further be explored to develop an understanding of their roles in strong quantitative basal resistance in kiwifruit, and this may assist in the development of high-throughput genetic markers. Our study will help advance the development of marker assisted selection of tetraploid yellow-fleshed Psa-resistant kiwifruit seedlings, following validation of the QTLs. The findings from the stab assay further inform us that the quantitative nature of this trait is affected by a complex of genetic, environmental and developmental factors, as well as the method of inoculation and the phenotypes scored for QTL mapping.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-0817/9/11/967/s1, Figure S1: Bait density across Red5 chromosomes. (a) shows number of baits designed on each chromosome and (b) shows density of baits over the physical distance for each of the chromosome, Figure S2: PCA plot for the genetic dataset of the population. Parents of the population are in purple (P1) and red (P2) circles with a single sample of low call rate. High-quality F1 individuals are within the green circle. Clustering of outliers from the F1 population can be seen in the blue circle whereas F1 samples with low quality are highlighted in the yellow., Figure S3: Minor Allele Frequency of the variants in the population. The plot shows Minor Allele Frequency (MAF) of the SNPs exhibited in the population, Figure S4: Marker dosage conversion in a tetraploid. A. For a tetraploid, 5 different marker dosages classes are possible (or say ploidy +1) for a bi-allelic marker which originates from the counts of alternate allele (a) in green and reference allele (A) in red. The types of genotype dosages from 0 (1st class for the A allele) to dosage 4 (5th class for the A allele) are presented. B. The re-coding of markers is an important process in defining the segregation type. Here the four different possible scenarios are presented to understand the markers with simplex (1) × nulliplex (0) segregation type. Different marker dosage classes in each segregation type (1 × 0, 1 × 4, 3 × 0 and 3 × 4), are converted to the most obvious or simplest form of their segregation behaviour in the population, 1 × 0, Figure S5: QTLs for control of the Stab_Psa_score in the population infected with Psa in the stab assay. Plots show the profile of two Quantitative Trait Loci (QTLs) for the Stab_Psa_score phenotype in the tetraploid A. chinensis population. The red dotted line on the QTL profile indicates the LOD threshold at alpha 0.05. Below the QTL profile is information on the source of the allele from the parental homologues. P1 homologues are represented as h1-h4 and P2 homologues as h5-h8. Green represents the positive contribution of the allele h1 from P1 on LG25 and the alleles from h4 and h8 from P1 and P2 respectively, on LG29, Figure S6: QTLs for the control of Stem necrosis and Leafspot in the population infected with Psa in the stab assay. Plots show the profile of two Quantitative Trait Loci (QTLs) for the Stem necrosis phenotype on LG21 and Leafspot on LG 16 in the tetraploid A. chinensis population. The red dotted line on the QTL profile indicates LOD threshold at α 0.05. Below the QTL profile is information on the source of the allele in the parental homologues. P1 homologues are represented as h1-h4 and P2 homologues as h5-h8. Green represents the positive contribution of the allele from h2 and h4 from P1 on LG21 for the control of Stem_necrosis and the alleles from h5 from P2 on LG16 for Leafspot. Figure S7. Designations and pedigree of the parents. The pedigree shows designation of each parent, sex and the names of grand parents and great grand parents. The origin of the great grand parents is from a cross that was performed between the accessions imported from China, Table S1. List of annotated gene models under the QTLs. The raw read files for Capture-seq data in all genotypes are divided into two sub-datasets i.e., R1 and R2 reads and is submitted in a public repository https://zenodo.org. The DOI for file "Rawreads_R1.fastq.gz" is 10.5281/zenodo.4285666 and DOI for the file "Rawreads_R2.fastq.gz" is 10.5281/zenodo.4287637.