Alcohol acyl transferase genes at a high-flavor intensity locus contribute to ester biosynthesis in kiwifruit

Abstract

Volatile esters are key compounds contributing to flavor intensity in commonly consumed fruits including apple (Malus domestica), strawberry (Fragaria spp.), and banana (Musa sapientum). In kiwifruit (Actinidia spp.), ethyl butanoate and other esters have been proposed to contribute fruity, sweet notes to commercial cultivars. Here, we investigated the genetic basis for ester production in Actinidia in an A. chinensis mapping population (AcMPO). A major quantitative trait loci for the production of multiple esters was identified at the high-flavor intensity (HiFI) locus on chromosome 20. This locus co-located with eight tandemly arrayed alcohol acyl transferase genes in the Red5 genome that were expressed in a ripening-specific fashion that corresponded with ester production. Biochemical characterization suggested two genes at the HiFI locus, alcohol acyl transferase 16-b/c (AT16-MPb/c), probably contributed most to the production of ethyl butanoate. A third gene, AT16-MPa, probably contributed more to hexyl butanoate and butyl hexanoate production, two esters that segregated in AcMPO. Sensory analysis of AcMPO indicated that fruit from segregating lines with high ester concentrations were more commonly described as being "fruity" as opposed to "beany". The downregulation of AT16-MPa-c by RNAi reduced ester production in ripe "Hort16A" fruit by >90%. Gas chromatography-olfactometry indicated the loss of the major "fruity" notes contributed by ethyl butanoate. A comparison of unimproved Actinidia germplasm with those of commercial cultivars indicated that the selection of fruit with high concentrations of alkyl esters (but not green note aldehydes) was probably an important selection trait in kiwifruit cultivation. Understanding ester production at the HiFI locus is a critical step toward maintaining and improving flavor intensity in kiwifruit.

Figure 1

QTL, HiFI locus and phylogenetic analysis. A, Map of linkage group (LG) 3 from A. chinensis (Fraser et al., 2009) showing the position Ke714 and Ac2340 markers associated with the HiFI locus. B, Genomic organization of AATs at the HiFI locus on chromosome 20 from the Red5 kiwifruit genome assembly (Pilkington et al., 2018). AAT genes models are represented with an arrow. The genomic location, gene annotation, and size of the genes are listed in Supplemental Table S4. C, Phylogram of Actinidia AATs and previously characterized fruit AATs. Amino acid alignments were generated with Clustal W in Geneious (Version 10.0.3). Trees were inferred using the maximum likelihood method based on the JTT matrix-based model (Jones et al., 1992). Evolutionary analyses were conducted in MEGA6 (Tamura et al., 2013). Percentage bootstrap values (1,000 replicates) are given below each branch. Clades I and II AATs are boxed (Günther et al., 2011). MpAAT1 (apple, Malus domestica, KC291129), BanAAT1 (banana, Musa sapientum, AX025506), SAAT (strawberry, Fragaria ananassa, AAG13130), SpAAT1 (tomato, Solanum pennellii, KM975321), CmAAT1 (melon, Cucumis melo, CAA94432). From kiwifruit: A. chinensis var. deliciosa AdAT17 (HO772638), A. chinensis var. arguta AaAT18 (OM417333), A. chinensis var. chinensis AT16-MPa (HO772640), AT16-MPb (OM417331), AT16-MPc (OM417332), AcAT20 (OM417334), and AcAT24 (OM417335).

Figure 2

Expression of AATs at the HiFI locus and volatile ester production during development and ripening of Red5 kiwifruit. The x-axis shows samples collected from whole fruit at 30 days after anthesis (D) and from flesh collected during development from 60 to 150 D and at four ripening stages (H1–H4). A, Expression of AAT genes at the HiFI locus was measured by RT–qPCR using the universal primers listed in Supplemental Table S6. Ethylene production was measured as described previously (Atkinson et al., 2011). Data are mean ± se (n=3). Statistical analysis in GraphPad Prism: one-way ANOVA (analysis of variance) followed by Tukey’s multiple comparison test. For ethylene (dashed line), means with the same letter were not significantly different at the 0.05 level. For AAT expression (bars) only sample H4 was significantly different (**) to all the other samples at the 0.05 level. B and C, VOCs were collected using SPME and analyzed by GC–MS during fruit development and ripening. Only compounds found at >10 ng·g⁻¹ FW in at least one time point are presented. Data are mean ± se (n = 3). Compounds: 1, a branched 2-methylbutanoate ester; 2, (Z)-3-hexenyl acetate; 3, methyl nonanoate; 4, 2-propenyl butanoate; EtBu, ethyl butanoate; 5, methyl benzoate; MeBu, methyl butanoate; 6, butyl butanoate; 7, ethyl hexanoate; 8, ethyl benzoate; 9, methyl hexanoate; 10, ethyl propanoate. Complete volatile data for flesh and skin are given in Supplemental Tables S5, A and B.

Figure 3

Activity of AT16-MPa–c in N. benthamiana and E. coli. AT16-MPa–c were cloned in the pHEX2 binary vector, expressed transiently in N. benthamiana and assayed for their ability to catalyze production of ethyl butanoate (EtBu), butyl hexanoate (ButHx), and hexyl butanoate (HexBu). A, Plants were inoculated with pHEX2_GUS as a negative control, MpAAT1 (white bars) as a positive control, pHEX2_AT16-MPa (black bars), and pHEX2_AT16-MPb and c (gray bars). The enzyme with the highest efficiency was set at 100% for each ester compound. Data are mean ± se (n = 4). B–D, Recombinant AT16-MPa/c proteins were overexpressed in E. coli and assessed for AAT activity using the optimal conditions described in Supplemental Figure S4. Data are mean ± se (n = 3). B, AT16-MPa/c activities with butanoyl CoA and a range of alcohols as substrates. C, AT16-MPa/c activities with butanol and a range of CoAs as substrates. D, AT16-MPa activity with hexanoyl CoA and a range of alcohols as substrates (left) and with hexanol and a range of CoAs as substrates (right).

Figure 4

Volatile production in transgenic AT16-MPa–c downregulated and control kiwifruit lines. Kiwifruit were treated with ethylene at 100 ppm for 24 h to ensure even and complete ripening. VOCs were extracted from outer pericarp tissue into solvent and analyzed by GC–MS. Data were collected in Year 1 from a control line and AT16-MPa–c transgenic lines T3501 and T3503. In Year 2, data were from a control and AT16-MPa–c transgenic lines T3499 and T3501. Total ester (black bars), aldehyde (gray), and alcohol (white) concentrations are presented. Individual volatile compound data are given in Supplemental Table S8. Data are presented as mean ± se (n = 3). Statistical analysis in GraphPad Prism: one-way (analysis of variance) ANOVA followed by Bonferroni’s multiple comparison test. *Significantly different from the respective control at P <0.01.

Figure 5

Contribution of esters to the sensory properties of kiwifruit. Ripe fruit from AcMPO were assessed for five variables: soluble sugar content (%SSC) and four sensory attributes (sweetness, acidity, texture, and overall sensory score). Single or multiple descriptors of fruit flavor were obtained in a free response text box and analyzed for term frequency vs high- and low-ester class (from HRM maternal marker analysis, Table 3) using TF–IDF. Principal components analysis was then performed on a matrix containing 53 sensory descriptors and the five variables. The axis of each variable (arrows) is projected against the principal component axes. Descriptors associated with the high-ester class are given as circles and with the low-ester class as triangles. The amount of variation (Dim%) explained by the first two dimensions is shown.

Figure 6

Variation in key volatiles between Actinidia domesticated and germplasm material. VOCs were collected by purge and trap then analyzed by GC–MS from ripe fruit of improved/domesticated “D” and unimproved/germplasm “G” vines of A. chinensis var. chinensis (A) and A. chinensis var. deliciosa (B). Log concentrations (ng·g⁻¹ FW) of two key volatile esters: methyl butanoate (MeBu) and ethyl butanoate, (EtBu) produced from the HiFI locus, and two key aldehydes: (E)-2-hexenal (E2-hex) and hexanal, are presented as box and whisker plots. Boxes show interquartile range, whiskers show 10–90 percentiles, and dots are data points outside this range. Statistical analysis in GraphPad Prism: one-way ANOVA (analysis of variance) followed by Bonferroni’s multiple comparison test with “D” sample compared to “G” sample for each compound. *Significantly different at P <0.05.