Field experiments.

Tomato genotypes for flavor analysis were grown from 2003 to 2010 in 12 field trials in Florida that encompassed spring (June) and fall (December) harvest seasons mostly at the Gulf Coast Research & Education Center (GCREC) on the central west coast (Wimauma) and winter harvest seasons (March) at the Tropical Research & Education Center (TREC) in south Florida (Homestead). Fruit of two genotypes, ‘Ugly Ripe’ and Cluster [also called tomato on the vine (TOV)] were bought at the table-ripe stage at a supermarket. Four December-harvest trials were done in 2003–05, and 2008; six June-harvest trials were done in 2004, 2005, and 2007–10; and three March-harvest trials were done in 2005, 2007, and 2008 (Table 1). The 2003 and 2004 trials were conducted at Bradenton except in Fall 2004 when the tomatoes were grown on a grower farm within 20 km of GCREC. The GCREC Bradenton soil type was classified as Eau Gallie fine sand with 1% organic matter and a pH of 6.5, thereafter the GCREC trials were conducted at Balm where the soil is classified as a Myakka fine sand siliceous hyperthermic Oxyaquic Alorthod with 1.5% organic matter and a pH of 7.3. The soil type at TREC was a Krome very gravelly loam with a high pH (7.8). The plants in the field were grown under standard Florida practices on fumigated soil under plastic-covered beds, and they were staked and tied (Olson et al., 2010). GCREC and the grower field plots were arranged in randomized complete block designs with two blocks and five to eight plants per plot. TREC trials were random single plots of 10 plants. Fruit at the table-ripe stage [U.S. Department of Agriculture stage 6, red, and for these samples, with 100% of surface having red color, and calculated on days past pollination for a genotype based on breeder experience (USDA, 2015)] were harvested the day before or on the morning of sensory panel analysis for all trials. All harvested fruit were stored at 24 °C in a laboratory until sensory panels were conducted. In the replicated trials, fruit were harvested and bulked from plants of both blocks for each genotype. Seed for December-harvest trials were sown in greenhouses from 20 July to 4 Aug., field transplanting was from 28 July to 13 Aug., and sensory panels were conducted from 5 to 19 Dec. Seed for June-harvest trials were sown in greenhouses from 28 Jan. to 16 Feb., field transplanting was from 11 to 25 Mar., and sensory panels were conducted from 7 to 14 June. Seed for March-harvest trials were sown in greenhouses from 5 to 10 Oct., field transplanting was from 10 to 14 Nov., and sensory panels were conducted from 8 to 11 Mar. Temperature, rainfall, and solar radiation data for spring and fall seasons at GCREC were presented by Georgelis et al. (2006). Within genotype, table-ripe, defect-free fruit with uniform size and color were used for both sensory and chemical analysis generally within 1 d of harvest as mentioned (held at ambient temperature until tasted and homogenized for chemical analyses). Sensory, SS, TA, individual sugars, and color analyses were conducted on all harvests in 2003–10. Individual acid (malate and citrate) analyses were done for 2007–10 trials and individual aroma volatiles from 2003–05 and 2007–10 (missing for 2006). A total of 38 genotypes including breeding lines, experimental hybrids, and commercial hybrid cultivars were tested in 13 sensory trials (Tables 1 and 3). Each harvest contained five to eight cultivars/genotypes. ‘Fla. 8153’ (Tasti-Lee^®) was tested in 12 out of the 13 sensory trials, and ‘Florida 47’ was tested in 11 out of 13 sensory trials. These two genotypes were then specifically compared because ‘Fla. 8153’ had scored higher for overall flavor than ‘Florida 47’ in six of seven panels (Scott et al., 2008). Other genotype/cultivars were tested in one to four sensory trials.

Table 1.Brief description of each tomato genotype used in this study of sensory and chemical flavor and the year and season (month) tested.

View Fullscreen

Sensory analysis.

The tomato fruit were washed, stem scar removed, and cut into wedges (about eight wedges/fruit) from at least 12 fruit per genotype, mixed, and presented on paper plates to a consumer panel consisting of 23 to 58 panelists (aged 18–65 years, roughly 40% to 60% female, the balance being male) per session (Stone and Sidel, 1997) with a randomized alphabetical code (A–H) presented to the panelists, so that they tasted samples in a randomized order. Plates from five to eight genotypes were placed on a table, with mixed slices from four fruit, and panelists were handed a score sheet with the sample order randomized (some score sheets started with sample A, others on sample B or C, etc.), with which to rate the intensity of sweetness, sourness, and overall flavor on a 1 to 9 hedonic scale, where 1 was least sweet, sour, or flavorful and 9 was most sweet, sour, or flavorful. Panelists were asked to taste more than one wedge and to partake of water and unsalted crackers between samples to cleanse their pallets. Plates of fresh samples from another three to four fruit were put out three to four times during each panel. For chemical analysis, subsamples of ≈8 to 12 slices per genotype were taken from each of the three to four plates/genotype presented to the panel. Wedges were homogenized in a blender (Waring Products Corp., Torrington, CT) for 30 s, allowed to stand for 3 min and 25 mL removed to a smaller blender with 10 mL room temperature-saturated CaCl₂, blended for 10 s, pipetted into 10-mL gas chromatography (GC) vials and flash frozen in liquid nitrogen for volatile samples and transported back in Styrofoam container with liquid nitrogen (Baldwin et al., 2011). The rest of the homogenate was put in a separate container on ice for transport to the USDA laboratory, where they were stored at −20 °C for sugar/acid/color samples and the GC vials at −80 °C for volatile samples until analysis. This resulted in three to four composite replications for sugar, acid, volatile, and color measurements per genotype from the same fruit presented to the panelists.

Sugar analyses.

The SS analysis was determined using a refractometer (RX-5000; Atago USA, Bellevue, WA), which was later replaced with a different model (PR-101; Atago, Tokyo, Japan). For analysis of individual sugars and acids, ≈40 g of juice was extracted using 70 mL of 80% ethanol solution. The mixture was boiled for 15 min, cooled and filtered with filter paper (Whatman No. 4; Whatman International, Maidstone, UK). The filtered solution was brought to 100 mL with 80% ethanol, and 10 mL was then filtered through a solid-phase extraction column (Sep-Pak C-18; Waters, Milford, MA) followed by a filter (0.45 µm; Millipore, Watford, UK). Individual sugar analysis was performed by high-performance liquid chromatography (HPLC) with a pump (SpectraSYSTEM P4000; Thermo Separation Products, San Jose, CA) and a refractive index detector (1100 Series; Agilent, Santa Clara, CA) equipped with a 300 × 6.5-mm column (Sugar-Pak, Waters) (Baldwin et al., 1991, 1998, 2004) and quantified on a standard curve. The mobile phase was 10⁻⁴ m ethylenediaminetetraacetic acid disodium calcium salt (CaEDTA) with a 0.5 mL·min⁻¹ flow rate at 90 °C. To better represent the sweetening power of the sugars, individual glucose and fructose concentrations were converted to sucrose equivalents [SE (Koehler and Kays, 1991)] by multiplying their concentrations by 0.74 and 1.73, respectively (Baldwin et al., 1998; Maul et al., 2000). Levels of glucose and fructose were summed to comprise total sugars (TS). Sugars were quantified using a standard curve (Baldwin et al., 1991).

Acid analyses.

Titratable acidity was determined by titrating to pH 8.1 with 0.1 n NaOH using an autotitrator (DL50; Mettler Toledo, Columbus, OH) and expressed as citric acid equivalents. The pH was also determined during this process. Organic acids (citrate and malate) were analyzed using the same HPLC as above equipped with an organic acid column (OA-1000 Prevail; Alltech, Flemington, NJ) with a flow rate of 0.2 mL·min⁻¹ at 35 °C and a mobile phase of 0.01 n H₂SO₄ and an ultraviolet detector (SpectraSYSTEM ultraviolet 6000 LP, Thermo Separation Products). The injection volume was 20 µL using an autosampler (Series 200; Perkin Elmer, Norwalk, CT). Acids were quantified using a standard curve (Baldwin et al., 2010).

Aroma volatile analyses.

Tomato volatile compounds were identified and quantified using head space analysis by gas chromatography [GC (model 8500, Perkin Elmer)] as in previous studies (Baldwin et al., 1991, 1998, 2004; Maul et al., 2000) using a headspace-sampler (HS-6, Perkin Elmer) equipped with a polar column [0.53 mm × 30 m, 1.0 µm film thickness (Durowax; J&W Scientific, Folsom, CA or Stabilwax; Resteck Corp., Bellefonte, PA)] and a flame ionization detector (FID), and in later years after purchasing an new GC system in 2006, using a different GC (model 6890, Agilent) (Bai et al., 2011) equipped with a polar (Stabilwax) and a nonpolar (low bleed) column [0.53 mm × 30 m, 1.5 µm film thickness (HP-5, Agilent)], where the flow was split equally between the two columns, and the detector was an FID. Since direct headspace volatile samples were not readily detectable by mass spectrometry (MS) (Song et al., 1997), confirmation by MS was accomplished by either a solvent extraction with ethyl ether as reported by Baldwin et al. (1991) for direct injection (for 2003–05 samples) or solid-phase micro extraction [SPME (50/30 µm DVB/CAR/PDMS; Supelco, Bellefonte, PA)] method (for 2007–10 samples) as reported by Wang et al. (2015). The instruments and settings were: for solvent direct injection: GC-MS (model 5890 GC + model 5970B MS; Hewlett-Packard, Norwalk, CT) with a 50-m wide-bore (0.31–0.32 mm film thickness) fused silica column of cross-linked 5% phenylmethyl silicone. For the SPME injection: GC-MS (6890 GC + 5973N MS; Agilent) with a nonpolar column [0.25 mm × 60 m, 0.50 µm film thickness (DB-5, Agilent)]. Volatile compounds were identified by the comparison of retention indices (RI) and mass spectra with library entries (NIST/EPA/NIH Mass Spectral Library version 2.0; National Institute of Standards and Technology, Gaithersburg, MD). The GC peaks for the aroma volatile compounds were quantified in microliters per liter using standard curves as determined by enrichment of bland tomato homogenate for both GC systems (Bai et al., 2011; Baldwin et al., 1991, 2004).

Fruit color analysis.

Fresh homogenate was measured in a cuvette for color using a chromameter (CR-300; Minolta, Tokyo, Japan) calibrated to a white plate using the CIE L*, a*, b* system including chroma and hue (Baldwin et al., 1991).

Statistical analyses.

SAS (version 9.3; SAS Institute, Cary, NC) was used for data analysis. Analysis of variance (ANOVA) was performed by using the general linear model (GLM) procedure for all individual measurements, including sensory panel, color values, sugar and acid contents, and volatiles where 23 to 58 individual panelists (replicates) were used in sensory panel attributes, and three to four replicates for all other chemical/physical measurements. For the analyses of each attribute, a randomized incomplete block split-split plot model was used with genotype (line) as the main plot factor, season as the subplot factor, and year as the sub-subplot factor. The main effects and the two-way interactions of each factor were included in the model. Multiple regression analysis (PROC REG) with two predictor variables was used to determine enhancement or suppression effects of individual volatile components to sweetness, sourness, or flavor. Cluster analysis (PROC CLUSTER) was used to test the distances between genotypes. All correlation analyses were conducted by using PROC CORR. For examination of the difference between the two most tested cultivars, FL 47 and Fla. 8153, the two independent-sample t test (PROC TTEST) was used. Principal component analysis (PCA) was conducted by using XLSTAT (Addinsoft, New York, NY) with standardized data.

Sensory analysis.

There were differences for sweetness by genotype; sourness by genotype and season, and interactions for genotype by season, genotype by year, and season by year; and for overall flavor ratings by genotype, season, and year, and a genotype by year interaction (Table 2). Sweetness scores ranged from 3.1 to 5.6, sourness from 2.9 to 5.3, and flavor from 3.3 (store-bought cluster tomato) to 5.7 (genotypes Fla.: 8630, 8153, 7907, and 5836) (Table 3). Also rated low in flavor were store-bought ‘Ugly Ripe’ and field-grown Fla. 8787, Fla. 8735, Fla. 8413, and industry standard ‘Florida 47’. ‘Fla. 8153’, later trademarked as Tasti Lee^® (Scott et al., 2008), was rated in almost every panel over the 7 years as was industry standard ‘Florida 47’ (Table 1). Comparing these two cultivars, ‘Fla. 8153’ rated relatively high for sweetness (5.1), intermediate for sourness (4.3), and was in the highest group for flavor (5.7) compared with the industry standard, ‘Florida 47’ (3.8, 4.1, and 4.1 for sweetness, sourness, and flavor, respectively), averaged over 414 and 384 panelists, for ‘Fla. 8153’ and ‘Florida 47’, respectively.

Table 2.F values and significant levels from ANOVA on sweetness, sourness and overall flavor of 38 tomato genotypes over 7 years, and one to three harvest seasons per year.^z

View Fullscreen

Table 3.Sensory ratings for 38 tomato genotypes over 7 years with one to three harvest seasons per year.^z

View Fullscreen

Sugars and acids.

There were differences for genotype, season, and year as well as genotype by season, genotype by year, and season by year interactions for SS, TS, glucose, fructose, SE, TA, SS/TA, citrate, malate (interaction for genotype by season only), and pH (Table 4). It is interesting that chemical sugar data were significant for season and year, while the sensory data were not (Table 2). This is likely due to the higher variation in the sensory than in the chemical data. Comparing just ‘Florida 47’ and ‘Fla. 8153’, ‘Florida 47’ had lower SS, TS, glucose, fructose, SE, SS/TA ratio, and pH while having higher TA, citrate, and malate (Table 5). This suggests that ‘Fla. 8153’ was generally sweeter than ‘Florida 47’, agreeing with the sensory ratings for sweetness, although the sensory panel did not rate ‘Florida 47’ higher than ‘Fla. 8153’ in sourness as the generally higher levels of acids for ‘Florida 47’ would indicate.

Table 4.F values and significant levels from ANOVA on sugar, acid, and color values of 38 tomato genotypes over 7 years, and one to three harvest seasons per year.^z

View Fullscreen

Table 5.Comparison of quality attributes between ‘Fla. 8153’ and ‘Florida 47’ tomatoes over 7 years, and one to three harvest seasons per year.^z

View Fullscreen

Internal color.

There were differences for chromameter “L” (lightness), “a” (green to red color), “b” (blue to yellow color), chroma (brightness), and hue (lower values indicate red color) for genotypes, seasons, years, and all interactions between genotypes, seasons, and years (Table 4). Comparing ‘Florida 47’ and ‘Fla. 8153’, ‘Florida 47’ had higher L, lower a, and chroma and higher hue values (indicating less red color). There was no difference between these cultivars for b values (Table 5). This is not surprising as ‘Fla. 8153’ is a high lycopene (red color pigment) cultivar (Scott et al., 2008) as a result of the crimson (og^c) gene (Thompson et al., 1965).

Aroma volatiles.

For volatile research, only 29 genotypes listed in Fig. 1 were used. Twenty-nine volatiles were analyzed over the years and seasons (Table 6) including acetaldehyde, hexanal, trans-2-pentenal, cis-3-hexanal, trans-2-hexenal, 2 + 3-methylbutanal, trans-2-heptenal, phenylacetaldehyde, methional, benzaldehyde, citral, trans-2, trans-4-decadienal, acetone, 1-penten-3-one, 1-octen-3-one, 6-methyl-5-hepten-2-one, β-damascenone, geranylacetone, β-ionone, 2-furanone, methanol, ethanol, 2 + 3-methylbutanol, cis-3-hexenol, 2-isobutylthiazole, linalool, 2-phenylethanol, methylsalicylate, and 1-nitro-2-phenylethane. There were differences between genotypes, seasons, years, and genotype by season interactions for 27 volatiles, except for citral (no difference for season) and β-ionone (only differences for genotype) (Table 6). There were also interactions for genotype by year and season by year for 18 volatiles (the 11 exceptions included hexanal, trans-2-pentenal, cis-3-hexanal, 2 + 3-methylbutanal, benzaldehyde, citral, β-ionone, 2-furanone, ethanol, cis-3-hexenol, and 1-nitro-2-phenylethane, for one or both interactions). Comparing ‘Florida 47’ and ‘Fla. 8153’ (Table 7), ‘Fla. 8153’ had higher levels of nine volatiles (acetaldehyde, 2 + 3-methylbutanal, citral, methional, 6-methyl-5-hepten-2-one, geranylacetone, β-ionone, 2 + 3-methylbutanol, and linalool), while ‘Florida 47’ had higher levels of four compounds, all C-6 aldehydes and one corresponding C-6 alcohol (cis-3-hexenal, trans-2-hexenal, trans-2-heptenal, and cis-3-hexenol). From this information, it would seem that ‘Florida 47’ had higher levels of “green” notes (Table 7) denoted by the C-6 aldehydes and alcohols, and ‘Fla. 8153’ had higher levels of fruity floral notes from acetaldehyde, citral, 6-methyl-5-hepten-2-one, geranylacetone, β-ionone, and linalool as determined from spiking or GC-olfactometry (GC-O) studies (Baldwin et al., 2004; Tandon et al., 2000, 2001). There were 16 volatiles that showed no difference between these two genotypes (Table 7).

View Figure

• Download as Powerpoint
• Download Figure

Table 6.F values and significant levels from analysis of variants (ANOVA) on 29 volatiles of 29 tomato genotypes over 6 years, and one to three harvest seasons per year.^z

View Fullscreen

Table 7.Concentration of volatiles emitted from 'Fla. 8153' and 'Florida 47' tomatoes tested over 6 years and one to three harvest seasons per year.^z

View Fullscreen

Comparing sensory and chemical data for the 29 genotypes over 6 years, with one to three seasons per year, showed correlations between chemical compounds and sensory perception of sweetness, sourness, and overall tomato flavor (Table 8). Many chemical compounds had significant and positive correlations with perception of sweetness (P ≥ 0.05) including sugar measurements, as expected (SS, TS, glucose, fructose, SE, SS/TA), and also many volatiles (acetaldehyde, hexanal, trans-2-hexenal, 1-penten-2-one, 6-methyl-5-hepten-2-one, geranylacetone, β-ionone, 2 + 3-methylbutanol, cis-3-hexenol, and methylsalicylate), although only SS, TS, glucose, fructose, and SE had coefficients (r values) greater than 0.4. Quite a few compounds showed significant correlations to perception of sourness including acid measurements (positively to TA and negatively to SS/TA ratio) and volatiles (acetaldehyde, hexanal, trans-2-pentenal, phenylacetaldehyde, benzaldehyde, genanylacetone, β-ionone, 2 + 3-methylbutanol, 2-isobutylthiazole, linalool, and methylsalicylate), with only TA and SS/TA showing an absolute value above 0.4. For linear relationships (all positive) to perception of overall tomato flavor, all the soluble compound measurements were significantly correlated (SS, TS, glucose, fructose, SE, TA, and SS/TA ratio) as well as many volatiles (acetaldehyde, hexanal, cis-3-hexenal, trans-2-hexenal, 1-penten-3-one, 6-methyl-5-hepten-2-one, geranylacetone, β-ionone, 2 + 3-methylbutanol, cis-3-hexenol, and methylsalicylate) with only SS and SS/TA ratio having r values greater than 0.4.

Table 8.Pearson correlation coefficient (linear relationship, Pearson’s r) between sensory hedonic ratings and flavor chemical contents for 29 tomato genotypes tested in one to three seasons over 6 years.^z

View Fullscreen

Multiple regression was then used to determine the contribution of aroma volatiles for enhancement or suppression of perception of sweetness, sourness, or overall tomato flavor in a 2-predictor model (Table 9). For perception of sweetness, SE was chosen as the companion variable for prediction of the perception of sweetness, since it encompasses both sugars weighted for sweetness and was highly correlated to sweetness (Table 8). Perception of sweetness was significantly (P = 0.002) enhanced by SE (Table 8) along with the volatiles hexanal, trans-2-hexenal, 1-penten-3-one, geranylacetone, 2 + 3-methylbutanol, cis-3-hexenol, 2-phenylethanol, and methylsalicylate at P < 0.05, in some agreement with Bartoshuk et al. (2013), while 2-phenylethanol suppressed perception of sweetness (P ≤ 0.05). The sourness model included TA, since it measures general acids (titratable acids) and was highly correlated (P = 0.0001) with sourness (Table 8), which along with trans-2-hexenal, geranylacetone, 2 + 3-methylbutanol, 2-isobutylthiazole, and methylsalicylate enhanced sourness perception, while methional and 6-methyl-5-hepten-2-one significantly (P < 0.05) suppressed perception of sourness. For overall tomato flavor, SS/TA was the chemical variable used, since it encompasses both sugars and acids and was highly correlated (P = 0.002) with flavor (Table 8), and along with acetaldehyde, hexanal, cis-3-hexenal, trans-2-hexenal, 1-penten-3-one, geranylacetone, ethanol, 2 + 3-methylbutanol, cis-3-hexenol, and methyl salicylate, enhanced perception of overall tomato flavor at P ≤ 0.05. If P ≤ 0.1 were to be used for significance, as is acceptable for sensory data (Meilgaard et al., 1991), then benzaldehyde, 6-methyl-5-hepten-2-one, and 2-isobutylthiazole also enhanced flavor perception. It is interesting that most of the volatiles that enhanced sweetness or sourness also enhanced overall flavor. Also interesting was that several of the volatiles that enhanced sweetness (with SE) also enhanced sourness (with TA) including trans-2-hexenal, geranylacetone, 2 + 3-methylbutanol, and methylsalicylate. Stevens et al. (1977) found that sweetness and sourness were related and concluded that a higher level of acidity was necessary for tomatoes to have consistent sweetness. Perhaps these volatiles enhanced acidity and thus indirectly enhanced sweetness. Table 7 shows descriptors for the various volatile compounds analyzed in this study, showing how their descriptors change depending on the matrix (water, alcohol, or deodorized tomato homogenate). Some of the volatiles that enhanced perception of sweetness are sweet or fruity in a tomato matrix (acetaldehyde, 2 + 3-methylbutanol, geranylacetone, and 1-penten-3-one). On the other hand, methylsalicylate has been described as wintergreen mint, and the C-6 aldehydes and alcohol are all described as green (hexanal, trans-2-hexenal, cis-3-hexenal, and cis-3-hexenol). The aroma volatiles that enhanced sourness included the sulfur volatile unique to tomato, 2-isobutylthiazole, and a C-6 aldehyde and alcohol (cis-3-hexenal and cis-3-hexenol, respectively) as well as methylsalicylate, which give green/mint notes along with the somewhat sweet/earthy 2 + 3-methylbutanol. The potato/earthy methional and sweet/floral 6-methy-5-hepten-2-one (in tomato matrix) appear to have detracted from sour perception.

Table 9.Multiple regression analysis for volatiles to determine the contribution of a volatile to enhance/suppress hedonic sensory ratings of overall flavor, sweetness, and sourness based on a two-predictor model. Samples were from 29 tomato genotypes tested in one to three seasons per year over 6 years.^z

View Fullscreen

To compare all the sensory and chemical data, PCA analysis was used, which explained 71.18% of the sensory and chemical data variation in two principal components (F1 and F2), and revealed that there were genotype and seasonal effects (Fig. 2). Dividing the figure into quadrants, quadrant A had relatively high loadings for component F2 and low loadings for F1, quadrant B had high loadings for components F1 and F2, quadrant C had low loadings for F1 and 2, while quadrant D had low loadings for component F2 and high for F1. Quadrants A and C were devoid of flavor descriptors and flavor chemical components including all volatiles due to low loading on F1. All of the ‘Fla. 8153’ samples and related genotypes through its parents (Fla. 7907 and Fla. 8059) were in B and D quadrants, associated with high loadings for sourness, sweetness, and flavor perception as well as all volatiles, (total) sugars (TS = total sugars: fructose, glucose), and moderate for SE and acid (TA), with higher pH in opposite quadrant measurements. Contrarily, almost all of the ‘Florida 47’, and several other low-scoring genotypes (‘Ugly Ripe’, Cluster, ‘Sanibel’, ‘Solar Fire’) were in quadrants A and C which had low loadings for most chemical and sensory measurements including sourness, sweetness, flavor, volatiles, sugars, or acids. Other than a few outliers (‘Florida 47’ June 2008 and 2009 and ‘Sanibel’ June 2010, which were just over into quadrant D), most of the March- and June-harvested samples were in quadrants C and D (mostly D) with high loadings on component F1 for flavor and sweetness perception as well as some volatiles and all sugar measurements (SS, glucose, fructose, SS/TA, sucrose equivalents, total sugars). December-harvested fruit (highlighted in blue) were mostly found in quadrant B, followed by A, with a few in C and D, with B having high loadings for some volatiles, sourness perception, and acid measurements (TA), opposite of higher pH (in the opposite quadrant) on components F1 and F2, and quadrant A being devoid of any sensory descriptors or chemical components.

View Figure

• Download as Powerpoint
• Download Figure

Cluster analysis was calculated based on all chemical and sensory data for the genotypes (Fig. 1). High flavor-scoring genotypes (Table 3), including ‘Fla. 8153’, parent Fla. 7907, and related genotypes clustered together along with ‘Fla. 8153’s other parent (Fla. 8059, midrange in flavor scores), while low flavor scoring genotypes, such as ‘Florida 91’ and ‘Florida 47’, were significantly separated from ‘Fla. 8153’ and related genotypes by cluster analysis. The group that had the largest distance units (4.1748) from the group that included ‘Fla. 8153’ included 9250, Cluster tomatoes, ‘Sanibel’, 8393, 8787, ‘Solar Fire’, ‘Ugly Ripe’, and ‘Florida 47’ (Fig. 2), which except for 9250 and 8393, generally had low flavor scores (Table 3).

When comparing only ‘Florida 47’ and ‘Fla. 8153’ using PCA (Fig. 3), 85.98% of the variation for all chemical and sensory data was explained in two PC components (F1 and F2). Quadrants B and D contained almost all the volatiles, most of the ‘Fla. 8153’ and June samples, and were associated with flavor and sweetness perception as well as (total) sugars and other sugar measurements (sugars = TS, glucose, fructose and SE). Quadrants A and C had most of the ‘Florida 47’ and December (highlighted in blue) samples (March samples being interspersed between December and June), and were associated with sourness perception and TA, as well as a lack of aroma volatiles. The few ‘Fla. 8153’ samples found in Quadrants A and C were all from December harvests.

View Figure

• Download as Powerpoint
• Download Figure

The findings of the study show that there is much yearly and seasonal variation for tomato flavor quality as seen by variation in sensory perception (Tables 2 and 3; Fig. 2) backed up by variation in levels of various chemical flavor components (Tables 4 and 6; Fig. 2). This is evident for all the genotypes investigated generally, and for the two cultivars with the most data, ‘Fla. 8153’ and ‘Florida 47’ (Tables 3, 5, and 7; Fig. 3). From these data, insight into seasonal effects and chemical drivers of tomato flavor can be gained.

The trials represent a diversity of growing environments. In general, spring crops (June harvest) go from shorter to longer days with increasing temperatures and solar radiation; fall crops (December harvest) go from longer to shorter days with decreasing temperatures and solar radiation; and winter production would tend to be under short days with relatively lower temperatures and solar radiation. Seasonal differences were evident, despite the yearly and seasonal variations, in that flavor chemical components and flavor perception (overall flavor and sweetness) were more associated with June followed by March and least associated with December harvest seasons (Figs. 2 and 3). This was confirmed by the fact that June harvests were more associated with sugars, followed by March and lastly December (Figs. 2 and 3). The precise cause of increased sugars in June is not known although solar radiation, daylength, and temperatures were all higher at that time of year (Table 10). This likely relates to increased photosynthesis resulting from increased solar radiation over long days. Georgelis et al. (2006) showed significant line × season (fall, spring) interaction in Florida for inheritance of sugars, with sugar levels being generally higher in spring for genotypes with genetically increased sugars, and suggested that this was due to higher solar radiation in spring than in fall. Early reports also showed that higher solar radiation was associated with higher sugars (Davies and Hobson, 1981; Winsor and Adams 1976). Our results do support the earlier findings, but possible effects of temperature and perhaps daylength could also play a role.

Table 10.Summary of weather data over the 2 weeks before harvest for the three tomato growing environments used in this study.

View Fullscreen

Despite the yearly and seasonal variation, drivers of tomato flavor were distinguishable and not surprisingly included sugars, acids, and aroma volatiles in general. In particular, the volatile aldehydes (especially green–note C-6 aldehydes and a related alcohol, cis-3-hexenol); fruity ketones and unique volatiles like the sulfur compound, 2-isobutylthiazole; the sweet-note alcohol, 2 + 3-methylbutanol; and the citrusy terpene alcohol, linalool and the terpinoid aldehyde, citral (which comprises the cis-/trans-lemony isomers of geranial and neral) seemed to play important roles. Tieman et al. (2012) had similar findings relating aroma volatiles to consumer preferences. Differences between the more flavorful ‘Fla. 8153’ and ‘Florida 47’ were also evident over the noise of seasons and years. The lower flavor ratings for ‘Florida 47’ compared with ‘Fla. 8153’ may have been caused by lower sugars and higher acids, but also aroma volatiles with the former having a lower sugar/acid ratio, more green notes (C-5 and C-6 aldehydes and related alcohol), and fewer fruity/floral notes (Table 7), mostly ketones (6-methyl-5-hepten-2-one, geranylacetone, β-ionone); other aldehydes (acetaldehyde); and alcohol (2 + 3-methylbutanol) and terpenes (citral and linalool). However, ‘Fla. 8153’ also had higher sweetness ratings and sugar levels, which are highly correlated to overall flavor (Table 8) (Malundo et al., 1995). It is therefore difficult to determine the effect of the aroma volatiles alone. A strong point within this dataset suggests that the volatiles as measured over so many seasons and locations, delivering significant differences, rising above the environmental noise, probably relate to the flavor differences in the two cultivars. Thus, the data provide volatiles and volatile levels for tomato breeders to target in their flavor improvement work. Further work is needed to verify the importance of this suite of volatiles, which will not be a trivial task; however, these data offer a manageable number of volatiles on which to focus. In addition, more was learned in this study about flavor perception. Many of these same volatiles were positively correlated with perception of sweetness (acetaldehyde, 1-penten-3-one, 6-methyl-5-hepten-2-one, geranylyacetone, β-ionone) or enhanced sweetness in combination with sugars (1-penten-3-one, geranylacetone), or enhanced overall flavor in combination with the sweet/sour SS/TA ratio (acetaldehyde, hexanal, cis-3-hexenal, trans-2-hexenal, benzaldehyde, 1-penten-3-one, 6-methyl-5-hepten-2-one, geranylacetone, ethanol, 2 + 3-methylbutanol, cis-3-hexenol, 2-isobutylthiazole, and methylsalicylate) (Table 9). In summary, it appears that tomato flavor quality is based on a balanced volatile profile (less green and more fruity volatiles) with moderate acid levels and relatively high levels of sugars.