Ten tomato genotypes with different fruit color were selected as parents. Genotypes used in the crosses are described below:

1. “NC1CS” is a large-fruited fresh-market tomato breeding line with the “crimson” (og^c) gene. It has large, smooth fruits and flesh is dark red in ripe fruits [21] .

2. “NC74CAP (2009)-Bk” has blue skin color, and small round fruited. Plant has determinate growth habit.

3. “NC72LYC (2009)-Bk” has yellow color with the β-gene.

4. “NC08133-1(x)-7W” has yellow and red stripes on the skin.

5. “NC28L-1(2008)” has large fruits with red color.

6. “NC95LYC (2009)-Bk” has large fruits with (r-mutant) yellow color.

7. “NC1Y” has yellow fruits with the “tangerine” (t) gene.

8. “Chocolate Stripe” has jumbo fascinated fruits with brown-purple stripes. This is an heirloom variety with indeterminate growth habit.

9. “Malachite Box” also has jumbo fascinated fruits with green flesh color. It has indeterminate growth habit.

10. “NC161L” [22] has pink fruit color and determinate growth habit.

All parents were crossed in diallel combinations without reciprocals to produce 45 F₁ hybrids in the fall season of 2009. Four experiments were conducted including 45 F₁ hybrids and 10 parents in the field (summer of 2010 and 2011) and in the greenhouse (2010 spring and 2011 fall) at the Mountain Horticultural Crops Research and Extension Center, Mills River, NC,

Seeds were planted in 72 cell flats (56 × 28 cm²) in potting mix and transplants, at about 6 weeks from seed were planted by hand in the field for summer experiments, and in the plastic pot for greenhouse experiments. Plants of each genotype were planted with 45 cm of plant-to-plant, and 150 cm row-to-row spacing in the field for summer experiments. The soil was a clay-loam and the natural day light photoperiod was about 14/10 hr with 25˚C - 30˚C high and 14˚C - 16˚C low temperatures. Similar temperatures and photoperiods were maintained in the greenhouse.

An experimental unit consisted of four plants in the greenhouse experiments and six plants in the field experiments. They were arranged in a randomized complete block design (RCBD) with three replications in each experiment. Parents and their F_1s were planted in a three gallon (11356.24 cm³) plastic pot using potting mix in the greenhouse. Fruits were harvested when they were ripe.

Fully ripe fruit (USDA stage 6) were harvested from the middle four plants from each line. Fruit were taken to the horticulture laboratory immediately after harvest and rinsed with tap water to remove debris and then air dried. A minimum of six fruit per genotype, free of injury, were sliced in half longitudinally (from stem to blossom end) with a sharp knife.

Tomato Analyzer (TA) software [ver. 2.2 (http://www.oardc.ohio-state.edu/vanderknaap/tomato_analyzer.htm)] [23] was used to obtain color parameters CIE L^*, a^*, b^*, hue and chroma estimates. Images were taken from cut fruits using a flatbed scanner (Canon Image Scanner CanoScan 8800F Canon, Melville, NY). The scanner and fruit were covered with a cardboard box with a black background to avoid effects of ambient light. Detailed instruction on taking images of tomato fruits is described by Rodríguez, Moyseenko [24] . Images of at least six fruit per line were taken and saved as “.JPEG” file. The Tomato Analyzer was used to estimate the color parameters L^*, a^* and b^* of the image. Average data of all fruit per scan was entered in an Excel file for further analysis.

Following the above measurements, about 250 - 400 g of tomato (one-half slice per fruit) was placed in plastic zip lock bags and held at −80˚C until analyzed for lycopene. Tomato slices were thawed and pureed using a homogenizer (Polytron PT10-35-GT with 20 mm PTDA generator, Kinematica, Bohemia, NY). Diluted purees (1:4 w/w water) were mixed in 50 mL centrifuge tubes then 20 mL poured into a 10 cm cuvette. Absorbance at 350 nm to 760 nm was measured using a Hunter Lab Ultra Scan Pro equipped with diffuse/8 xenon illumination (Hunter Associates, Reston, VA). Lycopene was calculated by subtracting the peak area at 750 nm (background light scatter) from that at 560 nm (the peak absorbance wavelength for total lycopene). Total lycopene was calculated using the system of Davis et al. [25] . Lycopene values were verified using HPLC and spectrophotometer analyses according to Perkins-Veazie, Collins [26] (data not shown).

Analysis of variance (ANOVA) of lycopene and color parameters (L, a^*, b^*, hue and chroma) of F₁ and parents for each experiment and over experiments were carried out separately using SAS software [27] . The ANOVA model over experiments assumed genotypes to be the fixed effect and experiments to be random. Genotypic differences among experiments were separated by least significant difference (LSD) at 5% probability level. Combining ability analyses of F_1s were conducted separately according to Griffing [28] by using method 4 (partial diallel including parents) and model 1 (fixed effect) as described in Diallel-05 [29] to estimate the general (GCA) and specific combining ability (SCA) effects. Variance components were also estimated by the same models.

Among the parents, Chocolate Stripe (74.9 mg/kg) was highest for lycopene content followed by NC 1CS (69.3 mg/kg) and NC161L (64.6 mg/kg). Level of lycopene in Chocolate Stripe was the same across the environments indicating that it was producing in the similar levels of lycopene in each experiment. NC1CS (dark red) and NC 161L (pink) were also one of the top lycopene containing parent lines. Dark red color in NC 1CS is because of the crimson gene, which gives the gel an orange color. The pink color of NC161L is due to red flesh and a clear (transparent) skin. High lycopene content in the pink colored tomato indicated that probably the lycopene was being measured from flesh itself. Yellow color genotypes such as 08133-1(×)-7W and NC 1Y had much less lycopene as expected. Another yellow color genotype NC95LYC (2009) had no detectable lycopene. Levels of lycopene of different parents are given in Table 1.

Chocolate Stripe was the parent with a purple brown stripe and a high lycopene content, Chocolate Stripe × NC 161L produced better hybrids (70.2 mg/kg) than Chocolate Stripe × Malachite Box (63.8 mg/kg) since NC 161L was better than Malachite Box for lycopene. Similarly, NC 1CS × Chocolate Stripe hybrid was better (66.1 mg/kg) than 08133-1(×)-7Wx Chocolate Stripe (60.3 mg/kg) hybrid for lycopene level since the latter cross involved a yellow fruited parent (Table 1). All crosses with Chocolate Stripe had the best level of lycopene in every environment. This indicated that Chocolate Stripe was one of the parents with best GCA.

Red fruited parents were NC 1CS and 28L-1 (2008). NC 1CS combined extremely well with all other parents including red, yellow, blue, pink, green and a mixture of purple and brown colored tomato lines for lycopene content as well as color parameters. Even with the yellow parents, lycopene was present when crossed with yellow and green parents (Table 1). Hybrids involving yellow parents including 95LYC (2009), 08133-1(×)-7W and NC 1Y produced reduced levels of lycopene. For example, NC 1CS × 08133-1(×)-7W hybrid had 61.4 mg/kg whereas NC 1CS × NC 1Y had 60.4 and NC 1CS × 95LYC (2009) had 56.9 mg/kg lycopene, respectively. This indicated that lycopene producing genes were dominant over yellow colored genes in tomato. Similar to NC 1CS, hybrids with 28L-1 (2008) produced good levels of lycopene even with yellow colored hybrids (Table 1). This indicated that red color was dominant over other colors, which was reflected in lycopene content.

Pink fruited parent, NC 161L, was one of the best parents for lycopene content. It produced high level of lycopene when combined with all colored parents including yellow. The best performing hybrid for lycopene was derived from Chocolate Stripe × NC 161L in this study (Table 1). In the pink fruited tomato, the skin is clear (transparent) and flesh is red. This indicated that pink fruited tomato is dominant for lycopene content in F₁.

Blue tomatoes are believed to have higher flavonoid levels rather than lycopene. In fact, lycopene level was towards the low end in this study in blue tomato parent (74CAP (2009)). Hybrids developed with this parent produced relatively low lycopene content except the one with 74CAP (2009) × Chocolate Stripe (69.5 mg/kg). This indicated that although less influential in lycopene biosynthesis, blue colored tomato was dominant over yellow (at least 48 mg/kg lycopene) and green (59.5 mg/kg lycopene) colored tomatoes. However, purple-brown stripe was dominant over blue since the former level was higher over blue tomato in the F₁ (Table 1).

There were four yellow colored parents with different gene combinations in this study namely 72LYC (2009) with the β-gene, 08133-1(×)-7W with red stripe, 95LYC (2009) with ther-gene, and NC1Y with the tangerine (t) gene. Several interesting hybrid combinations were found for lycopene content. Although both are yellow, hybrid of 72LYC (2009) × 08133-1(×)-7W had almost normal level of lycopene content (51.7 mg/kg). Reduced levels of lycopene were found in the hybrids developed from 72LYC (2009) × 95LYC (2009) (39.4 mg/kg), 72LYC (2009) × NC 1Y (41.7 mg/kg), 08133-1(×)-7W × NC 1Y (50.6 mg/kg) and 95LYC (2009) × NC 1Y (52.2 mg/kg). However, 08133-1(×)-7W × 95LYC (2009) produced extremely low levels of lycopene (6.8 mg/kg). Furthermore, both of these parents when crossed with Malachite Box either produced extremely low lycopene or

did not produce at all (Table 1). This indicated that yellow color was dominant over green suppressing lycopene content production.

Malachite Box was the only green flesh parent in this study. When crossed with other colored parents, lycopene content was found to be influenced by the gene of other parent (Table 1). This indicated that green flesh color was recessive to almost all of the colors in this study.

There was a significant difference (p < 0.05) among genotypes for lycopene content and color parameters including L^*-value, a^*-value, b^*-value, hue and chroma (Table 2). However, there was no interaction (p > 0.05) between genotypes and environments for lycopene content indicating that the level of lycopene was consistent across the experiments, and lycopene content is not influenced by environment.

Parents were found to be significantly different for general combining ability (GCA). Based on the GCA, the best parent with positive contribution to lycopene content was Chocolate Stripe followed by NC 1CS, 28L-1 (2008) and NC 161L, respectively. Other parents including Malachite Box, NC 1Y, 72LYC (2009), 08133-

1(×)-7W and 74CAP (2009) were the negative contributors to the lycopene content in the hybrids, respectively (Table 3). There was no difference among the parents for specific combining ability (SCA). Although statistically it was non-significant, a few combinations were notable. All hybrids with Chocolate Stripe and NC 1CS had positive gain for lycopene content whereas those with 28L-1(2008) and 74CAP (2009) had a mixture of positive and negative contribution with different parents. Other parents either did not have any contribution at all or there was a negative contribution for lycopene content. There was no GCA x environment and SCA x environment interaction for lycopene content (Table 3).

In order to estimate the heritability from the variance components, parameters were estimated as described by Zhang, Kang [29] . Narrow-sense heritability for lycopene was only 9% whereas broad-sense heritability estimate was 25%. Narrow-sense and broad-sense heritability estimates for L^*-value, a^*-value, b^*-value, hue and chroma were close to that of lycopene (Table 4).

When lycopene data were examined against color parameters, there was a negative correlation with L^*-value (r = −0.77) and b^*-value (r = −0.79) whereas it was positive with a^*-value (r = 0.67) and hue (r = 0.89). There was no correlation at all between lycopene and chroma (r = 0.06). This pattern is presented in Figure 1. Generally L^*-value represents the lightness of the color which means lack of redness. Similarly, b^*-value means yellow light which means no lycopene, which are negatively correlated. On the other hand, a^*-value indicates the intensity of redness, which means there is more lycopene which is what we have found in the present study.