Ninety-two F₁ bi-parental hybrids obtained from a cross of Amrapali as a female parent and Sensation as a male parent along with the parents were used for the QTL mapping. These parents were selected based on their contrasting features for key fruit quality-related traits including fruit color and firmness. The hybridity of the progenies has been confirmed earlier using SSR markers (Ramachandra et al., 2021). The population was phenotyped for multiple years and shows considerable variation for different fruit quality traits ( Figure 1 ). This mapping population is conserved at the field repository of ICAR-Indian Agricultural Research Institute, New Delhi, India.

Genotyping of the mapping population and parents in duplicate was carried out using an Affymetrix mango genome 80K SNP genotyping chip array (MiSNPnks 96 array, Dr. N. K. Singh, unpublished) having a total of 18,816 genes belonging to five different categories of genes such as single-copy mango genes (SCM), conserved single-copy genes between citrus and mango (CSCCM), cloned horticulturally important mango genes (HIM), disease resistance defense response-like mango genes (DRDRM), and multi-copy genes (MCR). The average SNP density is ~6 per gene, which is highly significant in linkage mapping and QTL identification studies in mango.

DNA for genotyping was isolated from the young leaves of individual trees of the mapping population and parents following the CTAB method as modified by Ramachandra et al., 2021. Genomic DNA quality was checked by electrophoresis in 1% agarose gel and quantified using a nanodrop spectrophotometer (NanoDrop 2000 Spectrophotometer, Thermo Scientific, USA). For target probe preparation, 50 μl of genomic DNA with a concentration of 10 ng/ul was used according to Affymetrix Axiom^® 2.0 Assay Manual. The DNA samples were pre-amplified using Target Prep Protocol QSCB1 (P/N 702990), fragmented and hybridized on the chip, followed by single-base extension through DNA ligation and signal amplification. The Affymetrix GeneTitan^® platform was used for staining, washing, and scanning of the chip signals as per manufacturer’s protocol. SNP allele calling was done using Axiom™ Analysis Suite version 2.0 using its three workflows, i.e., best practices, sample QC, genotyping, and summary on the Affymetrix Gene Titan. The Axiom Analysis Suite requires stored library files to convert CEL files into genotype calls. SNPs with low call rate across the samples were removed, and only good quality SNPs with a DQC of >0.85 and call rates of >95% were used for further analysis.

Genotype calls from all SNP markers generated by the Affymetrix GeneTitan^® platform from 92 progenies and parents were appended into a single csv file for export to Excel. A total of 80,816 SNPs were amplified, out of which loci with >5% missing data were filtered out. Markers amplified in only one of the two replications of the parents were also removed. The SNP allelic patterns between parents were taken as a reference for allele assignment in the mapping population. Monomorphic SNP markers were removed because they would not be informative for finding recombination events. Further, homozygous SNP markers for different alleles between the parents (aa, bb, or vice versa) were also removed, as there would be no segregation for such markers in the F₁ population. Thus, a total of 32,916 informative polymorphic SNP markers were identified between Amrapali and Sensation ( Table 1 ).

Based on analysis of allelic variation between parents for each polymorphic locus, SNP markers were classified as per the expected Mendelian segregation ratio in the mapping population. This set of SNP markers were mapped on the Dashehari physical map (unpublished) to assign their location on the chromosomes, and only those which mapped back on the chromosomes were retained. Polymorphic SNPs identified between parents represented multiple SNPs per gene. Therefore, only single SNPs per gene were selected ( Tables 1 , 2 ).

Linkage analysis was performed using homozygous SNP markers in one parent and heterozygous in the other parent (lm × ll or nn × np) and heterozygous in both parents (hk × hk) using JoinMap version 4.1 (Kyazma, Wageningen, The Netherlands). Further, the selection of SNP markers was based on their disomic inheritance and chi-square test of goodness of fit (p >0.05). Initially, for the construction of individual female and male maps, 4,834 markers in lm × ll and hk × hk, and 3,964 markers in nn × np and hk × hk were used, respectively. The initial grouping of markers was based on the independence log of the odds (LOD) tests in step ranging from 1.0 to 7.0. Other parameters were set as default. The regression mapping algorithm and Kosambi’s mapping function with minimum LOD of 3.0 were used for calculating the marker’s order. Markers showing suspect linkages were excluded in phases. Integration of female and male maps in which genotypes of some or all loci were determined in both populations took place, and the data from the separate populations were combined to calculate the integrated map. The groups that related to the same LG with at least two loci in common were combined by using the combined groups from the map integration function of the Join menu. The recombination frequencies and LOD scores of the selected sets of loci were combined into a combined group node in the navigation tree.

Such a combined group node is identical to a group node of a pairwise data population. The map calculations are based on mean recombination frequencies and combined LOD scores. For each pair of loci, the numbers of recombinant and non-recombinant gametes in the individual populations were calculated from the estimated recombination frequencies and corresponding LOD scores. The total numbers of recombinant and non-recombinant gametes of overall populations were calculated by totaling the numbers of the individual populations. From this, the mean recombination frequency and the combined LOD score were obtained. For map integration, the regression mapping algorithm was used.

Matured fruits from the F₁ trees of Amrapali/Sensation population and parents were carefully harvested with a 2 cm pedicel portion. The fruits were selected from all direction of the tree for evaluating fruit color and firmness traits for two consecutive years, 2019 and 2020. Fruits were washed thoroughly to remove the adhering dirt and dust, and rolled over the blotting paper to remove extra moisture on the surface and air dried. Fruit maturity was determined after harvest, and fruits having a specific gravity of ~1.01 to 1.02 were selected. These fruits were then wrapped in kite paper and placed in wooden boxes to ripen uniformly at room temperature. Initially, 30 fruits per individual genotype were subjected to ripening, of which 12 fruits showing uniform ripening and that were free from damage were selected for further analysis.

Peel and pulp color in terms of L*, a* and b* were measured using a Hunter-Lab Colorimeter (Model No. Miniscan^® XE plus 4500 L, Hunter Associates Laboratory, Inc., VA, USA). The instrument (45 / 0 geometry, D 65 optical sensor, 10 observer) was calibrated with black and white reference tiles through the tri-stimulus values X, Y and Z, taking as standard values those of the white background (X = 79.01; Y = 83.96; Z = 86.76) tile. Fruit peel color was measured at three points on the fruit surface, i.e., shoulder, middle, and bottom ( Figure 1 ). Mango pulp collected from ripe fruits was homogenized, and the color of homogenized pulp was measured using a ring and disk attachment.

The firmness of ripe mango fruits was measured with the help of the TA-XT Plus Texture Analyzer (Stable Micro Systems, UK). A 2.0 mm diameter stainless steel cylinder probe was used for the test in compression mode using the load cell of 5 kg capacity. Fruit firmness was expressed in Newton (N). Firmness was measured at three points (shoulder, middle, and bottom) of the fruits of each individual and parents ( Figure 1 ) with pre-test, test, and post-test speeds of 5, 2, and 10 mm/s, respectively (Jha et al., 2006; Jha et al., 2010). During the compression of mango fruit by the cylinder probe, firmness was determined by the highest force recorded in the force-time curve recorded by the Texture Analyzer. The first peak in the texture curve was taken as peel firmness. The average force between the first and second anchors was used to calculate the flesh firmness. Fruit color and firmness concerning individual progenies were observed under three replications having a minimum of three fruits per replication. The data on various parameters were subjected for Qstats analysis to know the basic quantitative statistics, viz., mean, variance, standard deviation, skewness, kurtosis, and average deviation. To test the normal distribution of traits in the mapping population, a test of normality was performed, and the critical values for rejection were 5.99 and 9.21 for the tests at the 5% and 9% levels of probability, respectively.

Phenotypic data of fruit color and firmness for the bi-parental F₁ population was used for QTL mapping with an integrated linkage map using MapQTL^®6 (Van Ooijen, 2009; Kyazma B.V.R, Wageningen, The Netherlands) using cross-pollinated (CP) for population type and Multiple QTL model-based MQM mapping for association statistics with mapping step size of 1 cM and regression function. Other calculation parameters were set with MapQTL default. The QTL statistics were reported for those in which the LOD score exceeded the threshold, and LOD peaks were used for determining the position of a significant QTL on chromosomes.

A total of 3,317 markers heterozygous in female (1:1) and 1,517 heterozygous in both parents (1:2:1) were used for the construction of the female map. Similarly, 2,447 heterozygous in male (1:1) and 1,517 heterozygous in both parents (1:2:1) were used for male linkage mapping.

Finally, two high-density individual linkage maps with 3,213 and 1,781 SNPs distributed on 20 LGs in each segregation category were constructed as female and male maps, respectively ( Supplementary Table 1 ).

The female map had 3,213 markers on 20 chromosomes with an average of 160.65 markers/chromosome. It covered a total map distance of 2,844.39 cM with individual chromosomes ranging from 82.41 cM (Chr 15) to 222.19 cM (Chr 3) with an average length of 142.22 cM. The average interval ranged from 0.43 cM (Chr 15) to 2.50 cM (Chr 3), with an average interval of 1.01 cM. The number of markers on each chromosome ranged from 75 (Chr 17) to 262 (Chr 4). This map is highly dense, and the density of SNPs ranged from 0.40 per cM (Chr 3) to 2.34 per cM (Chr 15) with an average of 1.21 SNPs per cM ( Table 3 ).

A total of 1,781 markers were successfully mapped on the 20 chromosomes in the male map with an average of 89.05 markers/chromosome. The male map covered a total map distance of 2,684.22 cM. The individual chromosome length ranged from 62.55 cM (Chr 15) to 187.86 cM (Chr 6), with an average length of 134.21 cM. The number of markers ranged from 42 (Chr 7) to 208 (Chr 11). The average interval ranged from 0.64 cM (Chr 15) to 3.68 cM (Chr 6), with an average of 1.74 cM. The number of SNPs per cM ranged from 0.27 (Chr 6) to 1.57 (Chr 15), with an average of 0.69 markers per cM ( Table 3 ).

Map integration was attempted using individual female and male linkage data, and a high-resolution integrated map comprising 4,361 markers mapped on the 20 chromosomes was constructed. Before the construction of integrated map, the LGs of individual maps were matched chromosome-wise, and a minimum of two markers common to individual maps were used for the construction of the integrated map. This map is highly dense, as the number of SNPs ranged from 108 (Chr 17) to 315 (Chr 11) with an average of 218.05 SNPs per chromosome. The integrated map covered a higher total map distance of 2,982.75 cM of mango genome compared to individual female and male maps. The individual chromosomes ranged from 85.78 cM (Chr 15) to 218.29 cM (Chr 4), with an average of 149.14 cM per chromosome. The average map interval on 20 chromosomes ranged from 0.34 cM (Chr 15) to 1.33 cM (Chr 3), with an average interval of 0.73 cM. The number of SNP markers per cM ranged from 0.75 for Chr 3 to 2.96 for Chr 15, with an average of 1.54 ( Table 3 ; Figure 2 ).

Phenotypic data on fruit color and firmness traits were generated for the 92 bi-parental F₁ population for two consecutive years in 2019 and 2020 ( Figures 3 – 5 ), and their mean value with genetic linkages observed in the integrated map was used for QTL mapping. Color coordinates a^* , b^* , and L^* indicate the red and yellow colors, and brightness on three points on the mango fruit in 2019, 2020, and the mean of two seasons were considered as individual traits. MQM mapping using MapQTL6 resulted in the identification of QTLs for 12 out of 20 phenotypic parameters analyzed for mapping with significant LOD scores in at least one season. Tables 4 – 6 show the 12 traits with significant LOD scores and their QTL position on the chromosomes.

Traditional mango breeding is cumbersome and time-consuming. In the present decade, rapid advancement in DNA sequencing and molecular genetic techniques has generated a wealth of information on mango genomics. Apart from conventional breeding approaches, efforts have been made in the recent past to utilize biotechnological tools for marker-aided breeding. Significant progress has been made in the area of genome sequencing of mango by several researchers (Singh et al., 2014; Luo et al., 2016; Singh et al., 2016; Kuhn et al., 2017; Wang et al., 2020). The wealth of genome resources has also been generated by our group as genome sequences of important mango cultivars (unpublished), and millions of SNPs have been identified, which has immense value in future mango breeding (Singh et al., 2014; Mahato et al., 2016; Singh et al., 2016; Iquebal et al., 2017).

As a key step in genetic linkage mapping, the mapping population is highly important. F₁, F₂, backcross populations, and haploid populations can be used for genetic mapping (Wang, 2001). The F₁ population is an ideal mapping population for highly heterozygous tree species like mango and can be obtained by one-generation hybridization between contrasting parents with high heterozygosity (Lu et al., 2019). There are several examples where genetic maps have been constructed based on the F₁ population such as Poncirus trifoliata (Xu et al., 2021), Ziziphus jujuba (Wang et al., 2019), Persea americana (Rendón-Anaya et al., 2019), Vitis (Zhu et al., 2018), Coryus avellana (Bhattarai and Mehlenbacher, 2017), and Theobroma cocoa (Argout et al., 2011). We chose to generate genetic linkage map using bi-parental F₁ population derived by crossing Amrapali/Sensation contrasting for fruit quality traits. Full-sib populations from two known parents are considered more effective for breeding progress than half-sib populations from open pollinated maternal parents. Genetic maps that are based on segregating full-sib hybrid populations are better for establishing linkages between horticultural traits and molecular markers for MAS (Kuhn et al., 2017).

Moreover, SNP marker-based genetic maps have several advantages, as SNPs are more abundant, easier to identify, easier to score, and unambiguous markers (Kuhn et al., 2017). Studies have shown that SNP loci are ubiquitous in the genome and the most abundant forms of genetic variation between individuals of the same species (Rafalski 2002; Kuhn et al., 2017). In this study, a mango 80K genic-SNP genotyping array was used to genotype the F₁ full-sib (Amrapali/Sensation) population along with parents.

High-density genetic maps are valuable in genetic and genomic studies, illuminating the genetic and molecular mechanisms of plants and providing the necessary framework for QTL analyses, gene cloning, and molecular breeding (Wang et al., 2019; Wu et al., 2019). We used 4,834 and 3,964 SNPs for construction of female and male maps. However, not all the SNPs that expected to segregate in a disomic fashion were able to be assigned to a linkage group. In female, 3,213 SNPs and in male 1,781 SNPs could be successfully mapped on 20 LGs. As the mango has 40 chromosomes with the haploid number of 20, we were successful in identifying 20 LGs for both female and male maps. This suggests the diploid nature of mango, and even though it is a partial allopolyploid, the two ancestral genomes were different enough to be distinguished by SNP markers (Kuhn et al., 2017). Genetic linkage data of female and male maps were used for integration, and a high-density linkage map was constructed, having 4,361 SNPs that covered the entire genome of mango. The map length covered by female, male, and integrated maps slightly differed for number of SNPs. The 3,213 SNPs of the female map covered 2,844.4 cM of the genome while 1,781 SNPs of the male map covered 2,684.2 cM. Integration of female and male maps resulted in 4,361 SNPs distribution across 20 chromosomes covering 2,982.8 cM of map length.

Few genetic maps in Mangifera indica have been constructed using different markers (Kashkush et al., 2001; Singh et al., 2014; Luo et al., 2016; Singh et al., 2016; Kuhn et al., 2017; Wang et al., 2020). The integrated genetic linkage map reported here is of high density, as the number of markers is reasonably high. In this study, we used a strategy to make the map that took advantage of the strengths of markers segregating differently. Further, markers corresponding to the same chromosome from all segregation groups were integrated. Our genetic map is significantly better than previous maps, as it is highly dense, based on full-sib (Amrapali/Sensation) population, and covered the entire mango genome.

Marker assisted selection provides ample opportunity to reduce the mango breeding cycle and improve efficiency as well. In recent decade, the advancements made in the augmentation of genome resources in mango (Kashkush et al., 2001; Luo et al., 2016; Singh et al., 2016; Kuhn et al., 2017; Wang et al., 2020) provide unprecedented opportunities to breeders for MAS in mango. Mango fruit is adored by people for its taste and nutrition, contributed by color, flavor, and aroma. Among these, peel and pulp color are important traits contributing to fruit quality and market value (Bajpai et al., 2017). Fruit firmness is another important trait for storage, transportation, and disease and insect management. Mapping populations from controlled crosses are not easy to generate in mango due to the high level of technical proficiency required (Bally et al., 2021). The F₁ bi-parental progeny population studied is few in fruit trees compared to annual crops (Grattapaglia and Sederoff, 1994). In the present study, we attempted to elucidate the regions of mango genome influencing fruit color and firmness using a F₁ bi-parental (Amrapali/Sensation) mapping population. To be useful for marker-aided breeding, it is imperative to have markers showing strong association with the horticultural traits. A map is not necessary to identify markers associated with a trait, but confidence in this association increases as multiple markers near the trait locus on the genetic map also show significant association with the trait (Kuhn et al., 2017). Fruit quality-related traits are quantitative traits that are influenced by multiple genes, and perhaps no single gene shows a significant impact on a trait. In our study, MQM mapping was used. This method has also been used in QTL analyses for quality traits in Gossypium hirsutum (Zhang et al., 2019), Poncirus trifoliata (Xu et al., 2021), and Elymus sibiricus (Zhang et al., 2019). In this study, we could identify SNP association for 12 of 20 traits used for analysis.

Fruit color is a highly variable trait within the fruit and is much influenced by environmental conditions. Precise qualitative evaluation of peel color is influenced by the number of fruits examined and from which part of the tree it is collected. More randomly chosen fruits from all parts of a tree may reduce this variation. In addition, we used another approach for determining the peel color using the Hunter color meter and took multiple observations on different positions of the fruit. This approach provided us with quantifiable data more suitable for QTL mapping compared to data based on scales or grouping. It was evident that red blush was more towards the shoulder compared to the middle and bottom portions of the fruit. An attractive fruit color is one of the most important factors for export markets (Nambi et al., 2016). The accumulation of pigments and their concentration and intensity determine the overall appearance of color in mango fruits (Pervaiz et al., 2017). Red coloration in fruits and other plant tissues has multifarious roles such as conferring plant disease resistance and protection against UV radiation (Bieza and Lois, 2001; Berardini et al., 2005; Sivankalyani et al., 2016). Anthocyanins also provide human health benefits against cancer and cardiovascular and other chronic diseases (Rao and Rao, 2007; Butelli et al., 2008; Singh et al., 2008). We found that fruit peel a* indicating red blush on the shoulder was associated with Chr 3. Four QTLs (AX-171381971, AX-171379053, AX-171375294, and AX-169929951) with high LOD value of 5.13-8.08 explaining around 35% of phenotypic variation were identified. The a* value observed on the middle of fruit resulted in associated SNPs located on Chr 2, 12, and 17. However, a* observed on the bottom portion of the fruit did not result in any QTL(s) in the present study. This may be because the appearance of red blush is more significant on the shoulder compared to the middle and bottom portion of the fruits. The SNPs associated with peel yellow color (b*) trait was on Chr 2, 14, 15, and 18. Mango fruits are classified based on peel color into green, yellow, and red types. Mango peel turns from green to yellow or red or retains green colors during ripening. Carotenoids and anthocyanins are the important pigments responsible for the colors of fruits. Pigments in different proportions may have an influence on the expression of different shades of color on mango peel. The genetics of peel color has not been studied in detail, but available reports indicate that it is governed by several genes and regulated in a more complex manner. In agreement with this, we observed several SNP markers housed on different chromosomes that had an influence on peel color development in mango. Consistency of these QTLs over phenotypic values observed in different seasons confirms their involvement in the expression of fruit color.

Mango varieties differ considerably for fruit peel and pulp firmness. Fruit firmness is a significant quality aspect of mango for consumers, as it represents ripeness and influences shelf-life, transportation, and processing issues (Jha et al., 2010). Various industries use puncture tests as part of their quality control procedure. Cool store operators monitor firmness throughout the storage period as part of their inventory management. In addition, firmness is sometimes used to predict consumer responses. Stec et al. (1989) found that preferred firmness at eating ripeness varied among assessors (Watkins and Harman, 1981; Harker et al., 1996). Jha et al. (2006) reported that the firmness of the mango fruits remained almost constant over the period of growth, and it decreased after attaining maturity. It was also suggested that the maturity of mango could be predicted by measuring size, color, and firmness. We used the approach of determining the peel and pulp firmness on different positions of fruits. We observed several SNPs associated with the peel and pulp firmness of mango fruits housed on Chr 3, 4, 6, 11, 19, and 20. SNPs on Chr 11 and Chr 20 consistently appeared in all seasons of observation and mean value of peel firmness. SNPs located on Chr 3, 4, 11, and 19 had an association with pulp firmness. SNPs located on Chr 11 at a position of 94.82 to 95.19 cM consistently appeared in different seasons. One QTL on Chr 3 at a position of 99.63 cM was also identified. The work reported here is unique concerning peel and pulp firmness results, where QTLs are reported for the first time in mango.