Background

Apples are a high-value food crop with unique nutritional qualities, stemming primarily from production of phenolic compounds. Flavonoids, a class of phenolics, contribute significantly to the color and nutritional value of apple, of which dihydrochalcones (DHCs) are a minor subclass. Contrasted with the diversity of phenolic compounds (thousands described), approximately 256 DHCs have been described in a wide distribution of plant species, though few are isolated from edible plants (Rivière 2016). Despite the rarity of DHCs, they are abundant in Malus species, and can represent up to 90% of the phenolic profile and 14% of the leaf dry weight (Gosch et al. 2010b). DHCs are absent or produced only in trace amounts in other food crops, including relatives within the Rosaceae. Phloridzin (phloretin 2′-O-glucoside) is the prominent dihydrochalcone in Malus species, including cultivated apple, and has anti-diabetic effects through increased insulin sensitivity and inhibition of sodium glucose cotransporter-2 (SGLT-2) (Najafian et al. 2012). Additionally, phloridzin has antioxidant, anti-cancer, anti-inflammatory, and phytoestrogenic effects (Puel et al. 2005; Nair et al. 2014). In a few Malus species, phloridzin content is replaced entirely by phloretin derivatives, sieboldin (3-hydroxyphloretin-4′-O-glucoside), and trilobatin (phloretin-4′-O-glucoside), which also have unique chemical properties. Higher antioxidant activity was observed in sieboldin relative to phloridzin and trilobatin, with a capacity to inhibit vasoconstriction and formation of advanced glycation end products associated with several degenerative diseases (Dugé de Bernonville et al. 2010; Xiao et al. 2017). Trilobatin, a sweet DHC, has anti-diabetic effects differing from phloridzin through α-glucosidase inhibition instead of SGLT-2 and reduction of chronic inflammation related to lipopolysaccharides (Dong et al. 2012; Fan et al. 2015). Trilobatin is about 300 times sweeter than sucrose, with potential as a naturally occurring commercial sweetener (Rivière 2016). Phloretin, the aglycone precursor to these DHCs, also has potential industrial value and nutritional quality, although it is found in lower concentrations (Behzad et al. 2017). Figure 1 depicts the synthesis of phloretin, phloridzin, sieboldin, and trilobatin.

Fig. 1
figure 1

Proposed synthesis of the dihydrochalcones phloretin, phloridzin, trilobatin, and sieboldin. Carbon double bond reductase (CDBR), chalone-3-hydroxylase (C3H), chalcone isomerase (CHI), chalcone synthase (CHS), flavonoid 3′-hydroxylase (F3′H), UPD-dependent glycosyltransferase (UGT)

Full size image

DHCs have been implicated in disease resistance in Malus, although there is no conclusive evidence directly relating DHC concentrations to resistance (Mikulič Petkovšek et al. 2008; Dugé de Bernonville et al. 2011). Fluctuations in DHC and flavonoid biosynthesis in response to infection were reported (Slatnar et al. 2012; Gaucher et al. 2013). Gosch et al. (2009) proposed that DHCs contribute to disease resistance through an oxidative cascade, and Gaucher et al. (2013) and Hutabarat et al. (2016) observed variation in this cascade in susceptible and resistant genotypes. Conversely, DHCs have been associated with apple replant disease by signaling and promoting pathogen growth (Yin et al. 2017). Phloridzin may act as an auxin regulator in Malus (Dare et al. 2017). Independent RNA interference silencing of chalcone synthase and phloretin-glycosyltransferases resulted in reduced auxin transport and aberrant tree architecture in apple (Dare and Hellens 2013; Dare et al. 2017). Whether or not sieboldin and trilobatin serve similar roles in Malus species needs to be researched.

While most Malus species produce phloridzin as the primary DHC, Malus toringo (Siebold) Siebold ex de Vriese and Malus sargentii Rehder replace phloridzin with sieboldin, and Malus trilobata (Poir.) Schneid. replaces phloridzin with trilobatin (Williams 1961). Interspecific hybridization with these species and other Malus species results in codominant inheritance of phloridzin, sieboldin, and trilobatin (Williams and Jarrett 1975; Hunter 1975). Williams and Jarrett (1975) first suggested DHCs in apple were controlled by independently segregating genes. Unlike other phenolics, with well-characterized metabolic networks, little is known about DHC synthesis and regulation. Its unique presence in cultivated apple has made apple somewhat of a model for DHCs. Most genetic studies in apple involve F1 progeny from single or multiple families (McClure et al. 2016). Advances in genotyping allow analysis of diverse germplasm through association studies, though this approach remains challenging in apple due to low linkage disequilibrium related to high heterozygosity and obligate outcrossing, and large populations required to detect small effect QTL (Kumar et al. 2013; McClure et al. 2014). QTL for several major apple phenolics have been reported, including a few minor, but no major QTL related to phloridzin content variation (Chagné et al. 2012; Khan et al. 2012; Verdu et al. 2014). More research is needed to understand the interactions of DHC metabolic pathways.

The aim of this study was to better characterize the genetic controllers of dihydrochalcone content in apple, particularly, those controlling sieboldin and trilobatin. We followed the segregation of major apple DHCs phloridzin, sieboldin, and trilobatin in five apple F1 populations and confirmed that DHCs segregate independently using chi-square analysis. Using linkage and association mapping, we identified marker-trait associations on apple linkage groups 7 and 8.

Materials and methods

Plant material

Five F1 populations were developed using Malus prunifolia (Willd.) Borkh. PI 589816 as the pollen parent. Seed parents included the ornamental cultivar ‘Evereste’ (population 16,705) and selections from the Cornell University apple breeding program in Geneva, NY; NY-152 (‘Wijcik McIntosh’ × ‘Suncrisp’|) (population 13,427), and three ‘Evereste’× ‘Red Jade’ progeny (populations 16,708, 16,709, and 16,710) (Table 1). Pedigrees are illustrated in Supplemental Fig. 1. M. prunifolia PI 589816, ‘Evereste’, and the ‘Evereste’ × ‘Red Jade’ progeny were selected based on dihydrochalcone profiles, containing phloridzin, sieboldin, and trilobatin. Population 13,427 was initially developed to study the columnar habit of NY-152. Each cross had high fruit set and seed viability (> 95%), although both PI 589816 and NY-152 are pale green lethal carriers and 25% of the initial seedlings died at the cotyledon stage in population 13,427 (Orcheski et al. 2015). Population 13,427 was from a cross made in 2013, germinated in the summer of 2015, overwintered in a greenhouse, and planted in spring of 2016. Leaf samples of 13,427 were collected from the greenhouse (fall 2015) and from the field (fall 2016). Populations 16,705, 16,708, 16,709, and 16,710 were developed in 2016, stratified in the winter of 2017, and planted in spring 2017. Samples were collected from the parents for HPLC and DNA extraction. Population 13,427 was evaluated for morphological traits including plant height, architecture, and internode length, all related to the columnar phenotype. A major QTL for columnar was identified previously on linkage group 10 (Morimoto and Banno 2015) and used to validate maps in population 13,427. PI 589816 and ‘Red Jade’ (PI 588786) are maintained on ‘EMLA 7’ rootstock in the USDA-ARS National Plant Germplasm System (NPGS) Malus collection in Geneva, NY. ‘Evereste’, NY-152, and ‘Evereste’ × ‘Red Jade’ selections are maintained on ‘Geneva 935’ rootstock. Leaf samples from 377 diverse accessions were collected from the NPGS Malus collection from 2013 to 2015 (Gutierrez et al. 2018).

Table 1 Pedigree, parental dihydrochalcone classes, and population size for each F1 population
Full size table

High performance liquid chromatography

Phenolics were extracted from mature leaf samples following the methods described by Gutierrez et al. (2018). Briefly, 25 mg of ground leaf tissue was suspended in 1.5 ml of 70% methanol (99.9%) with 2% formic acid (98–100%) for 10 min then centrifuged, with two technical replicates per genotype. Supernatant was filtered with a 0.20-μm syringe filter, and compounds were separated with an Inertsil ODS-3 column (4.6 × 250 mm; 5 μm, GL Sciences Inc., Tokyo, Japan) and 15% acetonitrile, 10% formic acid mobile phase at 1 mL/min on a Hitachi LaChrom Elite (San Jose, CA, USA) liquid chromatograph equipped with diode array detector. Standard curves for phloridzin, sieboldin, and trilobatin ranged from 7.8 μg to 1000 μg with R2 > 0.99, and sample values expressed in milligram per gram leaf fresh weight.

Genotyping

Individuals of population 13,427 were scored using genotyping-by-sequencing (Elshire et al. 2011) following the methods described previously (Gutierrez et al. 2018). Sequence tags were aligned to the M. domestica Whole Genome Reference Assembly v.2 (https://www.rosaceae.org/) using Bowtie 2 (Langmead and Salzberg 2012) with parameters D, R, N, L and i set to 30, 5, 1, 15, and S, 1, 0.25 to reduce misalignment. The Tassel 5 (Glaubitz et al. 2014) GBS pipeline was used for SNP calling. Ten replicates of parents NY-152 and PI 589816 were merged into single genotypes. VCFtools (Danecek et al. 2011) was used to filter data; twenty-six individuals were removed for low mean depth; sites were filtered for read depth of 8 and minor allele frequency of 0.20; and genotypes were filtered for 80% missing data. Kinship analysis and PCA identified twenty-five potentially outcrossed individuals which were removed from subsequent analyses. Genotypes were filtered based on Mendelian error calculated in PLINK 1.9 (Purcell et al. 2007) and set to missing. SNP markers were filtered based on chi-square tests for segregation distortion and > 0.95 similarity using JoinMap4.1. Genotypes from 280 of Malus germplasm accessions included in this study were scored previously (Gutierrez et al. 2018). Data included 274 individuals with mean depth above 18 (15th percentile) and 51,348 SNPs filtered for < 20% missing data, read depth of 10, minor allele frequency of 0.10, and QC > 98. LinkImpute (Money et al. 2015) was used to impute missing genotypes with parameters l and k set to 5 and 20.

Genetic maps and QTL analysis

GBS markers were divided into male (<nnxnp>) and female (<lmxll>) subsets based on JoinMap 4.1 (Kyazma, NL) conventions. Genetic maps for each parent were constructed with JoinMap 4.1, using regression mapping with linkages with recombination frequencies < 0.35 and LOD > 1.0, a jump threshold of 5.0 and Kosambi’s function. Loci were iteratively added to linkage groups based on strongest cross-link value. Interval mapping with regression analysis was done for each map separately using MapQTL 6 (Kyazma, NL) software. Significance threshold for each trait was determined through permutations test (n = 10,000). Association mapping was performed using the mixed linear model (MLM): Y = Xβ + Zu + e, where Y is a vector of DHC ratios; X and Z are design matrices; β is a fixed effects vector, which includes SNP markers, population structure (Q) determined by the first three principal components, and the model intercept; u is a vector of random additive genetic effects with covariance matrix 2 K\( {\upsigma}_{\mathrm{a}}^2 \), where K is a VanRaden kinship matrix (K) determined by SNP markers; X and Z are design matrices; and e are the residual effects (Wang et al. 2014). MLM was implemented in GAPIT (Lipka et al. 2012).

Data analysis

Principal component analysis and statistical tests, including chi-square tests for goodness of fit and independence, Pearson correlations, and ANOVA tests were calculated in R 3.3.2 (R Core Team 2016). Welch’s ANOVA and Games-Howell post hoc tests were used where sample size was unequal or equal variance could not be assumed, based on Bartlett’s and Levene’s tests for homogeneity of variance. Games-Howell tests were calculated in R package “userfriendlyscience” (Peters 2018). Mosaic plots and Pearson residuals were calculated using R package “vcd” (Zeileis et al. 2007).

Results

Dihydrochalcone content in F1 populations

DHCs were measured in all populations and their parents. Mean retention times across all samples for sieboldin, phloridzin, and trilobatin were 15.0, 20.3, and 30.7 min, respectively. Mean DHC content for parents of each population is presented in Table 2. Sieboldin was not observed independent of trilobatin in the progeny nor in the Malus germplasm (Gutierrez et al. 2018), though the proportion of sieboldin to trilobatin can be large with trilobatin detected in trace amounts.

Table 2 Mean leaf dihydrochalcone content (mg/g) in parents of apple F1 populations
Full size table

DHC genotypes of parents were inferred according to a three loci model and reported using codes A (phloridzin), B (trilobatin), and C (sieboldin) for simplicity. PI 589816 (SPT) is heterozygous for phloridzin, sieboldin, and trilobatin (AaBbCc), based on segregation in all five populations. NY-152 (P) is homozygous for phloridzin, and recessive for sieboldin, and trilobatin (AAbbcc), based on 1:2:1 segregation for PT, P, and SPT in population 13,427. ‘Evereste’ (SPT) is homozygous for phloridzin and sieboldin, and heterozygous for trilobatin (AABbCC), based on 3:1 segregation of SPT and P in population 16,705. In population 16,708, five DHC profiles were observed; P, T, PT, ST, and SPT, and selection 00121-005 was presumed to be heterozygous for phloridzin, sieboldin, and trilobatin (AaBbCc). Selections 00121-063 (SPT) and 00121-016 (SPT) are homozygous for phloridzin and heterozygous for sieboldin and trilobatin (AABbCc), based on the 9:4:3 ratio of SPT, P, and PT in populations 16,709 and 16,710. Progeny DHC profiles suggest they are controlled by three independently segregating loci (Williams and Jarrett 1975) as determined by a chi-square goodness of fit test (p > 0.05) (Fig. 2; Table 3). However, population 16,708 deviated significantly from expected ratios (χ2 (5) = 52.3, p < 0.001). Expected values were calculated from inferred genotypes, with the assumption that unobserved SP types of genotypes A_bbC_ manifest as P types.

Fig. 2
figure 2

Three-gene codominant model for phloridzin, sieboldin, and trilobatin segregation in population 13,427. Loci A, B, and C result in synthesis of phloridzin, trilobatin, and sieboldin, respectively, where sieboldin synthesis is dependent upon B allele (A_bbC_ produces only phloridzin). Loci presumed to be unlinked based on segregation ratios. NY-152 produces only phloridzin (P), and PI 589816 produces phloridzin, sieboldin, and trilobatin (SPT)

Full size image
Table 3 Number of offspring in each DHC profiles by population and Malus germplasm from Geneva, NY. Chi-square goodness of fit test based on inferred parental genotypes
Full size table

DHC content varied significantly within and among populations, based on Welch’s one-way analysis of variance and Games-Howell post hoc tests with Bonferroni correction. Phloridzin content (11.5 to 144.3 mg/g) was significantly different among populations (F(4,170) = 9.35, p < 0.001). Sieboldin content (3.1 to 57.9 mg/g) varied among populations (F(4, 77) = 55.98, p < 0.0001), and was significantly lower in population 13,427. Trilobatin content (0.1 to 46.5 mg/g) was different among populations (F(4, 103.9) = 138.6, p < 0.0001) and was significantly lower in population 16,705. Total DHC (15.8 to 144.3 mg/g) and varied significantly among populations (F(4, 6.9) = 34.4, p < 0.001). Mean and range of individual and total DHC content are listed by population in Table 4 and depicted in Supplemental Fig. 2. In population 13,427, mean DHC content was higher in 2016. This was likely due to environmental variation from greenhouse samples in 2015 to field samples in 2016.

Table 4 Mean and range of leaf dihydrochalcone content (mg/g) in apple F1 populations
Full size table

Individual DHC content varied by DHC profiles. Mean phloridzin content was higher in P profiles than in PT or SPT (F(2, 1284.5) = 423.56, p < 0.0001) and mean trilobatin content was higher in PT profiles than in SPT (t (1), p < 0.0001), consistent across all populations (Supplemental Fig. 3). Correlations among individual DHCs varied by population. In 13,427, there is a moderate correlation between sieboldin and phloridzin (r = 0.54), trilobatin and phloridzin (r = 0.52), and trilobatin and sieboldin (r = 0.50), where p < 0.0001. When subset by DHC profiles PT and SPT in population 13,427, the correlations between trilobatin and phloridzin are 0.74 and 0.71, respectively. Plotting trilobatin by phloridzin reveals two distinct groups based on dihydrochalcone profile (Supplemental Fig. 4). This pattern was not observed in other populations. There was a moderate correlation between phloridzin and sieboldin (r = 0.62, p < 0.001) in population 16,708 and between sieboldin and trilobatin (r = 0.52, p < 0.0001) in population 16,710. Other correlations among DHCs in these populations varied from weak to moderate but were not significant (p > 0.01).

The relative proportions of phloridzin, sieboldin, and trilobatin varied across populations. Phloridzin was the prominent DHC in all populations, with mean proportions > 60%. In population 13,427 phloridzin was the prominent DHC in all individuals, ranging from 64 to 87% of the total content in SPT profiles. A few individuals in the other populations had proportions of sieboldin and trilobatin > 50% and as high as 64%. Except for rare T profiles, trilobatin was generally a minor component in all five populations. The proportions of phloridzin, sieboldin, and trilobatin in progeny with PT and SPT profiles are presented in Table 5.

Table 5 Mean (range) of leaf dihydrochalcone proportions of apple F1 populations with PT and SPT profiles
Full size table

Genetic mapping

GBS produced 1,115,484 reads, 37.9% of which aligned uniquely to the reference genome, and 15.7% aligned > 1 time. After filtering, ~ 3100 and 1500 GBS markers were available for the male and female maps, respectively. Genetic maps were validated by mapping columnar architecture to linkage group (LG) 10 (Morimoto and Banno 2015) using plant height, internode length, and tree architecture descriptors. Interval mapping using the male genetic map identified a QTL on the bottom of LG 7 near marker LG7_27346079 at 67.1 cM, which explained 24.9, 22.4, and 50.2% of the variation of phloridzin, sieboldin, and trilobatin content, respectively. Another QTL on the bottom of LG 8 near marker LG8_33717934 at 63.7 cM explained 19.8% of the variation in sieboldin content. Male genetic maps of linkage groups 7 and 8 are highlighted in Fig. 3. Figure 4 illustrates the proportions of DHCs with the two most significant SNP markers associated with trilobatin and sieboldin content. Marker LG7_27743464 follows a 1:1 segregation between homozygous G and profile P, and heterozygous A/G and profiles PT and SPT. This marker is likely associated with trilobatin synthesis in this population, representing the B allele used in inferred parental genotypes. Marker LG8_31717934, associated with sieboldin synthesis, represents the C allele in inferred genotypes. Nearly all SPT profiles are heterozygous (C/T) at this locus, while most PT and over half P profiles are homozygous (C/C). About 40% of individuals of profile P also are heterozygous (C/T) for marker LG8_31717934, which suggests that predicted SP profiles manifest as P.

Fig. 3
figure 3

Male (M. prunifolia PI 589816) genetic maps of linkage groups 7 and 8 from population 13,427. SNP ID refers to chromosome (LG) and its physical position determined by sequence tag alignment to the ‘Golden Delicious’ Whole Genome Reference Assembly v.2

Full size image
Fig. 4
figure 4

Mosaic plot of proportions of dihydrochalcone (DHC) profiles by most significant SNP markers on linkage groups 7 and 8 in population 13,427. Box height represents proportion of individuals in DHC profiles and box width represents number of individuals of a genotype. Box colors represent deviation from expected observations assuming independence. Blue boxes have more individuals than expected, red boxes have fewer, and gray boxes are not significantly different based on Pearson residuals: \( {r}_{\mathrm{ij}}=\left({O}_{\mathrm{ij}}\hbox{--} {E}_{\mathrm{ij}}\right)/\sqrt{E_{\mathrm{ij}}} \)

Full size image

Within the Malus germplasm collection, we identified an association (Fig. 5) around the same position on LG 7 related to dihydrochalcone composition of 279 individuals. Marker LG7_27704634 homozygous G/G was tightly associated with profile P, whereas a significant portion of SPT, PT, and ST profiles was heterozygous G/T or homozygous T/T.

Fig. 5
figure 5

Manhattan plot based on Q+K model of leaf dihydrochalcone composition in Malus germplasm accessions. Negative log10 p values are plotted by physical position on 17 apple chromosomes. Horizontal line denotes Bonferroni adjusted significance threshold at α = 0.01

Full size image

Discussion

DHC profiles observed in these populations followed a pattern of three codominantly expressed, segregating loci. Chi-square tests for goodness of fit in all populations but 16,708 suggest these loci are unlinked. It is unclear why population 16,708 deviated significantly from the other populations. Selection 00121-005 was unique among the other ‘Evereste’ × ‘Red Jade’ selections, with reduced levels of sieboldin and trilobatin (Table 2). Additionally, it has unique morphological characteristics, including linear to lanceolate leaf shape, generally < 7-mm-wide and 28-mm-long leaves, shortened internodes (< 12.5 mm). Supplemental Fig. 5 highlights leaf variation between population founders. There are potentially other genetic factors affecting DHC concentrations in this population, which merit further investigation.

DHCs are structurally similar to narigenin chalcone (a precursor to flavonols, anthocyanins, and flavan-3-ols) but, deviate early through a Δ2–3 carbon double bond reduction of p-coumaroyl-CoA to p-dihydrocoumaroyl-CoA, which is converted to phloretin through chalcone synthase (Dare et al. 2013; Ibdah et al. 2014). In phloridzin synthesis, a molecule of glucose is added to 2′-position of phloretin via phloretin 2′-O-glucosyltransferases and other glycosyltransferases (Jugdé et al. 2008; Gosch et al. 2010a). Genetic analysis identified two major determinant loci on LGs 7 and 8 for trilobatin and sieboldin, though candidate genes were not identified near these loci. The biosynthesis of sieboldin has not been elucidated, but it may form through C-3 hydroxylation of phloretin, followed by glycosylation via phloretin 4′-O-glucosyltransferase (Hutabarat et al. 2016). Although the extent of pathway sharing between sieboldin and trilobatin requires further study, dependence of sieboldin on the determinant loci on LG 7 can explain why sieboldin is not detected independent of trilobatin. As population 13,427 did not segregate for phloridzin, no determinate locus for phloridzin was detected.

In addition to phloridzin, sieboldin, and trilobatin, other DHCs are reported in Malus. These minor DHCs, including phloretin, generally do not occur in appreciable quantities, raising questions concerning the preferential accumulation of phloridzin, sieboldin, and trilobatin. However, in these populations, we detected an unidentified peak around 13.1 min, with maximum absorbance at 280 nm. This peak was found in similar concentrations as sieboldin or trilobatin in these populations, was strongly associated with sieboldin, and interval mapping identified a QTL for this unknown peak at the same position on LG 8. This unknown could be a hydroxylated phloretin derivative, such as hydroxyphloridzin (3-hydroxyphloretin 2′-O-glucoside), but further investigation is needed.

Mean content for individual compounds varied by DHC profile, suggesting a partitioning effect among compounds. Mean phloridzin content was significantly higher in P profiles than in PT or SPT profiles, and mean trilobatin content was higher in PT profiles than in SPT profiles across all populations. Some significant positive linear relationships were observed between DHCs in some, but not all of the F1 populations, with correlations from 0.00 to 0.74. From our observations, the correlations are stronger between trilobatin and phloridzin content when subset into PT and SPT profiles. Additionally, we observed significant variation in the proportion of DHCs relative to their parental types. While, phloridzin was the major DHC across populations, sieboldin or trilobatin were dominant in several individuals.

Chemical analysis has focused on a few key points in the DHC biosynthetic pathway. A list of key genes investigated for dihydrochalcone synthesis both in vitro and in vivo is presented in Table 6. However, cause of DHC hyper-accumulation Malus is unknown. Studies have shown dihydrochalcone content can be regulated via enzymes outside its biosynthetic pathway (Khan et al. 2012; Rihani et al. 2017). The evolution of trilobatin and sieboldin in select Malus species is of particular interest. Trilobatin differs from phloridzin by the position of the glucose moiety, which could occur through a phloretin 4′-O-glucosyltransferase (UGT75L17) identified on LG 9 (Yahyaa et al. 2016). Interestingly, UGT75L17 was expressed in ‘Golden Delicious’, which does not accumulate trilobatin. Additionally, Dare et al. (2017) detected trilobatin accumulation after silencing phloretin 2′-O-glucosyltransferase UGT88F1 silenced in ‘Royal Gala’. These two reports highlight the biosynthetic potential for DHC synthesis beyond phloridzin in cultivated apple and emphasize the unknown mechanism for selective DHC accumulation.

Table 6 Genes related to dihydrochalcone biosynthesis
Full size table

The evolution of dihydrochalcone production in Malus is unique within the Rosaceae. Although phloretin and phloridzin are not detected in closely related pear (Pyrus communis), they can be synthesized from p-dihydrocoumaroyl-CoA with enzymatic extracts from pear leaves, suggesting biochemical divergence in Malus from pear at an earlier step (Gosch et al. 2009, 2010a). Apple-pear intergeneric hybrids are able to produce both phloridzin and arbutin, a pear-specific metabolite absent in apple (Fischer et al. 2014). Although phloridzin was reported in other rosaceous plants (rose and strawberry), it is not found in high concentrations (Gosch et al. 2010b). Phloretin is the precursor to each compound, though sieboldin may be formed following an additional hydroxylation step. Phloretin was not measured in this study and is generally a minor component of dihydrochalcone content or not reported in Malus phenolic studies involving various tissue types (Verdu et al. 2014; Tang et al. 2015; De Paepe et al. 2015). Presumably, phloretin is nearly completely glycosylated in planta as it is more reactive than its glycosylated forms, resulting in cytotoxicity (Gosch et al. 2009). Silencing of UGT88F1 in ‘Royal Gala’ reduced both phloridzin and phloretin, suggesting a feedback mechanism to inhibit the accumulation of phloretin in healthy plants (Dare et al. 2017).

There is evidence that phloridzin concentrations may affect other upstream pathways (Dare et al. 2017). Other QTL for DHCs have been reported in apple populations. Verdu et al. (2014) identified QTL on LGs 1 and 5 for phloridzin content in fruit and juice which explained 6.4 to 11.6% of the variation, and on LGs 3, 5, 12, and 15 for phloretin xyloglucoside content in fruit and juice which explained 17.4 to 23.6% of the variation. Chagné et al. (2012) identified a QTL for phloridzin xyloglucoside on LG 17 explaining 10.3% of the variation. No candidate genes were proposed for these reported QTL. This is the first report of DHC loci on LGs 7 and 8. Other traits mapped to LGs 7 and 8 include aphid resistance (Dysaphis devecta Walker and Eriosoma lanigerum Hausm.) (Cevik and King 2002; Bus et al. 2008); woolly apple aphid resistance allele Er3 from M. torigino ‘Aotea 1’ (Bus et al. 2008); powdery mildew from Malus hybrid ‘White Angel’ (Evans and James 2003); fire blight resistance from ‘Fiesta’ (Khan et al. 2007); fruit weight (Devoghalaere et al. 2012); and fruit acidity (Zhang et al. 2012).

Conclusions

Dihydrochalcones phloridzin, sieboldin, and trilobatin are unique compounds in Malus species. We observed significant DHC content variation among five F1 populations and segregation of DHCs in apple, following a model of three, independently segregating loci. Additionally, we identified the first genomic loci associated with trilobatin and sieboldin segregation in population 13,427 and Malus germplasm. Markers on LG 7 and LG 8 were strongly associated with trilobatin and sieboldin production, respectively. We posited that sieboldin is biochemically linked to trilobatin, based on the loci on LG 7. Loci identified in this study require further investigation but may facilitate the utilization of genetic resources to broaden the nutritional quality of new apple cultivars.