A taste of pepper: genetics, biochemistry and prediction of sweet pepper flavor

This PhD project started with the composition of a diverse panel of genotypes that represented, (i) roughly the flavor variation in the commercial <em>Capsicum annuum</em> breeding program of Rijk Zwaan, (ii) parents of available mapping populations and (iii) some genotypes that were expected to have extraordinary flavors. The complete set consisted of 35 genotypes of which 24 genotypes were non-pungent. Volatile and non-volatile compounds as well as some breeding parameters were measured in mature fruits of all genotypes throughout the growing season. In addition, from three harvests the non-pungent genotypes were evaluated for taste by a trained descriptive sensory panel. The biochemical profiling with use of SPME-GC-MS allowed visualization of between- and within-species volatile compound variation. Principal components analysis (PCA) on the intensity patterns of 391 putative volatile compounds revealed individual grouping of <em>C. chinense,</em> <em>C. baccatum </em>var. <em>pendulum</em> and <em>C. annuum</em>, indicating potentially interesting volatile variation present in the former two groups. A large group of saturated and unsaturated esters were mainly responsible for the individual grouping of the <em>C. chinense </em>accessions. Due to the elevated acid concentrations and aberrant volatile profiles of the <em>C. baccatum </em>var. <em>pendulum </em>accessions PEN45 and PEN79, the two BIL populations derived from these accessions were identified as interesting candidates for further study. Compared to e.g. Mazurka the citrate concentration of the <em>C. baccatum </em>accessions was 2.5-3 times higher and the malate concentrations were even up to 12 times higher (Chapter 2). Based on the non-pungent genotypes, we found highly correlated clusters of volatiles and non-volatiles, which could be related to metabolic pathways and common biochemical precursors (Chapter 3). Contrasts between genotypes were caused by both qualitative and quantitative differences in these metabolic clusters, with the phenolic derivatives, higher alkanes, sesquiterpenes and lipid derived volatiles forming the major determinants. For the description of the non-pungent genotypes the panelists used fourteen attributes to describe the flavor sensation in the mouth/throat, which were the texture attributes crunchiness, stickiness of the skin, toughness and juiciness, the basic taste attributes sweetness and sourness and the retronasal flavor attributes aroma intensity, grassiness, green bean, carrot, fruity/apple, perfume, petrochemical and musty. The variation in flavor could be reduced into two major sensory contrasts, which were a texture related contrast and the basic sweet-sour contrast. The structure of the PCA plots resulting from the analysis with one harvest (Chapter 3) and the analysis with the combined three harvests (Chapter 4) remained almost identical, indicating the stability of these contrasts. To relate the sensory attributes to the metabolite data and to determine the importance of the individual compounds we used Random Forest regression on the individual harvests and on the three harvests together. Several predictors for the attributes aroma, fruity/apple, sourness and sweetness were found in common between harvests, which we proposed as key-metabolites involved in flavor determination of sweet pepper (Chapter 4). This list contains compounds with known relations to attributes, like sweetness and sugars, but also several compounds with new relations. In this respect we have demonstrated for the first time, that the metabolites p-menth-1-en-9-al, (E)-β-ocimene, (Z)-2-penten-1-ol, and 1-methyl-1,4-cyclohexadiene are related to fruity/apple taste and/or sweetness of pepper. For sourness the only compound with a consistent significant contribution was an unknown C₆H₈O₂ compound. We postulated therefore the hypothesis that in pepper the effect of sourness related metabolites is masked by other volatile and non-volatile compounds or texture differences (Chapter 3). Subsequently in Chapter 4 we described a clear sweetness-sourness interaction and demonstrated that the masking effect of fructose and other sugars explained why we did not find organic acids contributing to the prediction of sourness. The major sensory attributes were also predicted between harvests. The Random Forest predictions of the texture related attributes (juiciness, toughness, crunchiness and stickiness of the skin) and sweetness were very good. The predictions of the attributes aroma intensity, sourness and fruity/apple were somewhat lower and more variable between harvests, especially in the second harvest. In general, we concluded that prediction of attributes with higher heritabilities works better and is more consistent over harvests (Chapter 4). Based on the results of the initial experiments (Chapter 2) the species <em>C. baccatum</em> was chosen for further study. To exploit the potential flavor wealth of <em>C. baccatum</em> PEN45 we combined interspecific crossing with embryo rescue, resulting in a multi-parent BC₂S₁ population, that was characterized for sensory and biochemical variation (Chapter 5). We developed a population specific genetic linkage map for QTL mapping of characterized traits. Because of the complex structure of our BC₂S₁mapping population we encountered several limitations, such as accidental co-segregation, underrepresentation of color linked markers and pre-selection leading to skewness, which might have resulted in false positive or missed QTLs. Despite these limitations, we were still fairly well able to map several biochemical, physical and sensory traits, as demonstrated at first for the (monogenic) control traits red color and pungency in the BC₂S₁ mapping population and in second instance by validation of genetic effects via an experiment with near-isogenic lines (NILs).This two-step approach turned out to be very powerful, since it led to the identification of the main results from this thesis: (i) Asmall <em>C. baccatum</em> LG3 introgression causing an extraordinary effect on flavor, which resulted in significantly higher scores for the attributes aroma, flowers, spices, celery and chives. In an attempt to identify the responsible biochemical compounds few consistently up- and down-regulated metabolites were detected, including the well-known pepper compound 2-isobutyl-3-methoxypyrazine (down) and 6-methyl-4-oxo-5-heptenal (up); (ii) Two introgressions (LG10.1 and LG1) had major effects on terpenoid content of mature fruits, affecting at least fifteen different monoterpenes; (iii) A second LG3 fragment resulted in a strong increase in Brix (total soluble solids) without negative effects on fruit size (Chapter 5). In Chapter 6 some extra sensory results of the pungent genotypes are given and a comparison between the two <em>C. baccatum pendulum</em> BILs (PEN45 and PEN79 derived) is made in light of the overall results. Finally the perspectives for breeding are discussed and presented in the form of a flowchart for flavor improvement.