Plant material

Two parents and the F₁ generation from the cross of the two genotypes were evaluated. The female parent (LCP-95) was the S₈ line derived from the native LOR95 accession with indeterminate growth habit and pepper-like fruit shape (local description), native to Tehuacan, Puebla. The male parent (Sh) was the LA1223 accession of S. habrochaites Knapp & Spooner native to Peru, with indeterminate growth habit and cherry fruit shape; this accession was provided by the Tomato Genetics Resource Center (TGRC) at the University of Davis, California, United States of America.

Phenotypic evaluation

The agronomic evaluation was carried out in the spring-summer 2016 and 2017 growing seasons, in a hydroponic greenhouse production system. Sowing for the 2016 cycle was conducted on March 12 in 200-cavity polystyrene trays. Peat moss (Kekkila^®) was used as substrate. The transplant was performed 43 days after sowing (das). Sowing of the second cycle was carried out on April 23, 2017 in conditions similar to those of the first cycle, and the transplant occurred at 40 das.

A randomized complete block experimental design with four replicates was used; the experimental unit consisted of 20 plants. Each plant was placed in a black 12 L polyethylene bag (12 L), with 4 mm volcanic sand as substrate to obtain a density of 2.7 plants·m^-2 in an open hydroponic system. Plants were spaced 35 cm apart and the rows 2 m apart.

Nutrition was provided with ^{Steiner's (1980)} universal solution at pH 5 and was applied according to the phenological stage: 0 to 40 days after transplanting (dat) at 50 %, and from 41 dat to the end of the cycle at 100 %. Plants were led to one main stem, which was blunted at 90 dat. Whitefly (Bemisia tabaci Gennadius) was controlled with imidacloprid and 50 lambdacialotrine + 100 chlorantraniliprole. In addition, carboxamide and metalaxyl-M + chlorothalonil were used to control Phytophthora infestans, and azoxystrobin to control Altenaria solani.

Fourteen agronomic and morphological traits were assessed in all plants in the experimental unit: days to flowering of the first cluster after transplant (DFFC), days to ripening of the first cluster after transplant (DR), number of flowers per cluster (NFL), number of fruits per cluster (NFR), total number of fruits per plant (TNF), number of fruit locules (NL), total fruit yield per plant (TFY, kg·plant^-1), average fruit weight of the third cluster (AFW, g), stem diameter (SD, mm), equatorial fruit diameter (ED, mm), polar fruit diameter (PD, mm), total soluble solids (mainly sucrose; B, °Brix), height of the first cluster (CH, cm) and total plant height (PH, cm).

Five fruits from the third cluster of each plant in the experimental unit were used to measure PD, ED, B and AFW. The measurement of CH and PH was made with a flexometer (model FH-3M, TRUPER^®, Mexico). TFY and AFW were evaluated with a digital balance (model SP2001, Ohaus^®, USA). B was measured with a digital refractometer (model PAL-1, ATAGO^®, Chian). For SD, PD and ED, a standard and metric digital vernier caliper (model CALDI-6MP, TRUPER^®, Mexico) was used.

Statistical analysis

Combined analysis of variance between assessment cycles and Tukey’s range test (P ≤ 0.05) were performed using SAS® statistical package ver. 9.0 (^{SAS Institute Inc., 2002}). The linear model used was Y _ijkl = µ + α _i + γ _j + δ _ij + B(C) + ε _ijkl ; where Y _ijkl is the observation of the i-th genotype in the j-th cycle of the k-th replicate within the l-th block, µ is the overall mean, α _i is the random effect of the i-th observation of the genotype, γ _j is the random effect of the j-th cycle, δ _ij is the random effect of the i-th genotype in the j-th cycle, B(C) is the random effect of the l-th block nested in the k-th replicate of the j-th environment and ε _ijkl is the random error associated with the experimental unit.

Heterosis with respect to the mid-parent was estimated with the formula: H=F1-MPMP×100; where H is percentage heterosis (%), F ₁ is the phenotypic mean of the F₁ population, MP is the mid-parent and MP is equal to Pi + Pj2 (Pi and Pj are the phenotypic mean of the parent i and j) (^{Falconer & Mackay, 1996}).

The average degree of dominance (DD) was calculated from the phenotypic mean of both parents (LCP-95 and Sh), the mid-parent (MP) and the F₁ generation_. Thus, it was considered as an additive (AD) action type when the F₁ mean was equal to MP, as partial dominance towards LCP-95 (PDL) if the F₁ mean was close to the female parent, as partial dominance of Sh (PDH) when the F₁ value was closer to the wild parent, and as overdominance (OD) only if the F₁ phenotypic mean was higher than its parents (^{Molina-Galán, 1992}).

For the calculation of narrow-sense heritability (h²), the mid-parent-offspring regression methodology was used (^{Falconer & Mackay, 1996}). This analysis was carried out with the RStudio 1.0 statistical package (^{Rstudio Team, 2015}), based on the simple linear regression model.

Analysis of variance

In the analysis of variance, between-cycle significance was detected for most variables, except for PD. Likewise, significance between genotypes was detected in all evaluated variables. This allowed identifying the existing variation among the evaluated genotypes. Genotype per cycle interaction was significant for TFY, AFW, NFL, DFFC, DR and NL (Table 1), which indicated that the genotypes presented consistency for most of their values across environments. The coefficients of variation suggested that experimental management and control were carried out adequately, which minimized the experimental error and gave validity to the results obtained.

Average degree of dominance and comparison between genotypes

According to the averages obtained for each variable (Table 2), the parents used had contrasting phenotypical traits. This suggests that by making a cross between the two progenitors, it would be possible to increase the genetic base for tomato improvement by generating germplasm with allelic combinations provided by the Sh material. Proof of this was that the plants of the LA1223 accession, compared to the LCP-95 parent, were more vigorous in height, but with thin stems, had more flowers and fruits but smaller size and a higher concentration of total soluble solids, and were later in flowering and fruit ripening.

From the comparison of means, the average degree of dominance of the evaluated traits was determined. In CH, NFL, DFFC and DR, partial dominance towards the LCP-95 parent was observed, because the F₁ genotype had lower average values compared to the mid-parent, but very close to LCP-95; in contrast, AFW, TFY, NFR and NL showed partial dominance towards the wild Sh parent, since the F₁ generation was similar to Sh (Table 2). Similarly, earliness and flower number traits were reported to be controlled by partial dominance effects towards wild tomato alleles (^{Hernández-Bautista et al., 2014}; ^{López et al., 2015}).

The traits ED and PD showed additivity, since the average values of the F₁ generation were similar to the mid-parent (Table 2). ^{Lippman and Tanksley (2001)} showed, in an interspecific cross of S. lycopersicum L. and S. pimpinellifolium L., that fruit size is regulated by genes with additive effects.

Overdominance in F₁ was identified in traits such as PH, SD, NTF and B, this because their values were higher than the best parent (Table 2). In different works, different results have been reported for total soluble solids and plant size, since these traits were influenced by gene action of dominance. This allowed inferring that Sh increases these traits to a greater extent, and that it has potential as a source of high-value alleles (^{Chattopadhyay, Dutta, Dutta, & Hazra, 2011}; ^{López et al., 2015}).

Heterosis

The average heterosis for both cycles ranged from -51 to 56 % in the evaluated traits. Of 14 traits measured, five had positive heterosis, namely PH, SD, B, NFR and TNF, and the rest showed negative heterosis (Table 3). Regardless of the type of value (positive or negative) of heterosis, this parameter will always be important for breeding. Positive heterosis is desirable to increase the magnitude of traits like fruit or grain yield; on the other hand, if it were negative, the magnitude of the traits would decrease (^{Alam et al., 2004}; ^{Escorcia-Gutiérrez, Molina-Galán, Castillo-González, & Mejía-Contreras, 2010}; ^{Martínez-Martínez et al., 2014}).

In the present work, it was observed that B and TNF had the highest positive heterosis, which indicated that Sh increased the values of these traits; in contrast, PD, ED, TFY and AFW presented the highest negative heterosis. This allowed inferring that Sh caused a decrease in fruit size and yield, so Sh would not be useful to improve fruit yield, but it would be useful in other fruit traits.

Commercially, S. habrochaites Knapp & Spooner is used for the formation of rootstocks employed in tomato production under adverse conditions, such as drought and low temperatures (^{Ntatsi et al., 2017}; ^{Poudyala et al., 2017}), so LA1223 could be used as a parent of hybrids to be used as rootstocks, or to derive lines that are outstanding for this trait and can be used per se as rootstock. This is because F₁ showed a 2 mm increase in stem diameter compared to the best parent. Regarding the impact of the wild parent on tomato yield, ^{Velasco-Alvarado et al. (2019)} showed that it increases yield compared to ungrafted plants and commercial rootstocks, and increases the amount of total soluble solids in fruit.

In total soluble solids, heterosis was observed to be high (49.20 %) and higher than other values reported in different studies. For example, ^{Hernández-Bautista et al. (2014)} used S. pimpinellifolium L. as a male parent and observed negative heterosis for total soluble solids (-0.98 %); in contrast, ^{Yadav, Singh, Baranwal, and Solankey (2013)} reported a hybrid of cultivated and wild tomato with high heterosis (37.50 %), but lower than that observed in this work. This indicates that Sh would be useful for improving internal fruit quality; however, it should be considered that using it also modifies traits such as fruit size, weight and shape.

The heterosis observed in NFR (14.30 %) is an intermediate value for the interval from -25 to 38.9 % heterosis with respect to the mid-parent reported by ^{Gul, Ur-Rahman, Khalil, Shah, and Ghafoor (2010)} for hybrids generated with wild accessions of S. lycopersicum. TNF showed high heterosis (55.90 %), which was found in the range observed by ^{Yadav et al. (2013)}, who reported heterosis of -52.50 to 58.5 % in intraspecific hybrids of cultivated and wild collections of S. lycopersicum L., although lower than that observed in an interspecific cross of S. lycopersicum L. and S. pimpinellifolium L. (183.93 %) (^{Hernández-Bautista et al., 2014}).

In precocity traits such as days to flowering and ripening, negative heterosis is important because it allows the reduction of these traits (^{Alam et al., 2004}). In this study it was observed that CH, DR and DFFC had negative heterosis, which allowed inferring that Sh decreased the days to flowering and ripening, cluster height and the number of fruits; however, it also reduces important traits for fruit size and weight, which makes the direct use of Sh for elite germplasm improvement difficult. This suggests the need to explore other breeding options such as backcrossing. ^{Martínez-Vázquez, Lobato-Ortiz, García-Zavala, and Reyes-López (2016)} report that for fruit size, weight and shape it is preferable to use native collections with similar traits to elite germplasm, since more than 50 % heterosis can be obtained for these traits.

Precocity traits were expressed in conjunction with unwanted traits such as lower weight and smaller fruit size, but it is possible to use Sh in breeding programs through different selection and line derivation techniques that conserve only the loci of the Sh parent that control the traits of interest, which could allow using the Sh alleles in an efficient way (^{Monforte & Tanksley, 2000}; ^{Zamir, 2001}).

Heritability

Average narrow-sense heritability was intermediate (44 %) (Table 2) according to the scale presented by ^{Molina-Galan (1992)}, ranging from 16 % (DR) to 94 % (TFP), which indicated that most of these traits are of additive effects. Thus, selection methods, such as line derivation, are an alternative for breeding in the case of some of the traits evaluated, while low-heritability traits would have to exploit non-additive effects by hybridization (^{Amaefula, Agbo, & Nwofia, 2014}; ^{Dagade et al., 2015}).

Traits such as TFY, TNF, SD, and PH had high heritability, indicating that environmental conditions had little impact on phenotypic variance. Thus, the heritability observed was the result of genetic differences between parents (^{Flint-Garcia et al., 2005}). ^{Hernández-Bautista et al. (2014)} and ^{Rodríguez et al. (2008)} reported similar values for inter- and intra-specific hybrids of S. lycopersicum L. and S. pimpinellifolium L. in the same traits.

The AFW, ED and B traits had intermediate heritability, so they could be used similarly to the high-heritability traits to take advantage of the additive effects (Table 2). ^{Parra-Gómez et al. (2016)} observed intermediate heritability for total soluble solids, fruit weight and fruit diameter in S₃ lines derived from an interspecific cross of S. lycopersicum L. with S. pimpinellifolium L.

Heritability was low in the traits NFR, NFL, PD, DFFC, DR, NL and CH. These values suggest that the observed variation was promoted by environmental factors and not by genetic factors, so the hybridization method would be the best for using these traits. ^{El-Gabry, Solieman, and Abido (2014)} identified in tomato hybrids that NFL, NFR, and DFFC are of low heritability, so they suggested that these traits are of non-additive effects.