Predicting Yield in Tomato Infected with Tomato Yellow Leaf Curl Virus (TYLCV) Using Regression Models Based on Physiological Traits

Abstract

Tomato yellow leaf curl virus (TYLCV) is one of the most destructive viral diseases causing severe yield losses in tomato production worldwide. This study investigated the effects of TYLCV infection on plant growth, photosynthetic physiological responses, and yield formation in greenhouse-grown tomatoes and evaluated the applicability of physiological trait-based yield prediction models. Two large-fruited tomato cultivars widely cultivated in Korean protected horticulture systems, ‘Daphnis’ and ‘Pink Star’, were inoculated with TYLCV under greenhouse conditions, and their growth, physiological responses, and yield characteristics were compared under high- and low-temperature growing seasons. TYLCV infection significantly reduced leaf length, leaf width, and leaf area index (LAI), and decreased both flowering truss number and fruit-setting truss number, resulting in reduced total yield. Physiological analyses showed that infected plants exhibited decreases in the OJIP fluorescence rise curve and Fv/Fm values, indicating a reduced photochemical efficiency in photosystem II. In addition, A–Ci response curve analysis revealed a reduction in net photosynthetic rate, suggesting limited carbon assimilation capacity. Total yield showed significant positive correlations with maximum net photosynthetic rate (Amax), Fv/Fm, and Ci300. GGE and GT biplot analyses further indicated that yield was closely associated with photosynthetic performance and canopy development traits. A multiple regression model based on physiological traits and virus infection status explained a significant proportion of the variation in tomato yield (R² = 0.367), indicating that TYLCV infection acts as a key limiting factor for yield reduction. These findings demonstrate that TYLCV infection restricts tomato productivity through reduced photosynthetic efficiency and altered canopy structure. Moreover, physiological trait-based yield prediction approaches may provide a useful framework for evaluating productivity under viral infection conditions and for developing data-driven crop management strategies in greenhouse tomato production systems.

1. Introduction

Tomato yellow leaf curl virus (TYLCV) is one of the most destructive viral diseases affecting tomato production worldwide. Infection with TYLCV causes yellowing and curling of young leaves, severe growth suppression, inhibition of flower truss formation, and substantial yield reduction [1,2]. The virus is primarily transmitted by the whitefly Bemisia tabaci, and under protected cultivation systems it can spread rapidly due to confined growing environments and continuous cropping practices. In Korea, greenhouse tomato production is commonly operated under year-round cultivation systems, which increases the persistent risk of TYLCV outbreaks. Once infection occurs, it can lead to significant economic losses due to reduced productivity and unstable crop performance [3,4].

The severity of TYLCV damage is strongly influenced by environmental conditions. Previous studies have shown that high-temperature environments tend to promote viral replication and symptom expression, resulting in more pronounced growth suppression and yield loss. Conversely, under lower-temperature conditions, symptom development may be mitigated or delayed [5,6]. These findings indicate that the impact of TYLCV infection should not be interpreted solely as the direct effect of a pathogen, but rather as a complex stress response arising from the interaction between viral infection and environmental factors, particularly temperature. Such interactions are especially relevant in greenhouse production systems where environmental conditions vary seasonally and can influence host–virus dynamics.

In Korean greenhouse tomato production, large-fruited red cultivars dominate the market due to their commercial value and consumer preference. These cultivars are widely cultivated with the aim of achieving stable yield under intensive production systems. However, even within the same cultivar, growth performance and yield response may vary considerably depending on annual climatic variation, differences in cultivation timing, and the stage at which virus infection occurs [7,8]. Previous TYLCV studies have mainly focused on comparing growth suppression and yield reduction between infected and non-infected plants. However, many of these studies were conducted under single-year experiments or restricted environmental conditions, limiting their ability to capture the combined effects of environmental variability and virus infection over multiple growing seasons.

In addition, relatively limited attention has been given to the physiological mechanisms underlying yield reduction caused by TYLCV infection. In tomato, photosynthetic performance, canopy light interception, and source–sink relationships are closely linked to fruit growth and yield formation, indicating that physiological traits can provide a mechanistic basis for explaining productivity variation [9,10]. Physiological indicators such as net photosynthetic rate and chlorophyll fluorescence are sensitive parameters that reflect plant stress status and photosynthetic efficiency. Despite their potential to quantitatively explain plant responses to viral infection, attempts to integrate these physiological traits into predictive models for yield variation remain scarce [11]. Although physiological and crop modeling approaches have been increasingly applied in tomato under non-stress or production-oriented conditions, including canopy photosynthesis analysis, whole-plant physiological regulation, and greenhouse crop modeling, comparable studies explicitly addressing pathogen-driven physiological constraints remain limited [12,13]. In particular, the development of yield prediction models based on physiological traits under conditions where biotic stress (virus infection) interacts with environmental stress factors, such as seasonal temperature variation, remains an important research challenge in greenhouse tomato production systems, particularly because existing tomato modeling studies have largely focused on non-infected crops or production optimization rather than pathogen-driven physiological constraints [14].

Therefore, the objective of this study was to investigate the effects of TYLCV infection and environmental variation on plant growth, physiological responses, and yield formation in greenhouse-grown tomatoes over two growing years in Chuncheon, Korea. Two widely cultivated large-fruited tomato cultivars were used for TYLCV inoculation experiments, and growth characteristics, photosynthetic responses, and yield traits were comprehensively evaluated under different seasonal temperature conditions. Furthermore, regression models based on key physiological traits, including photosynthetic parameters and chlorophyll fluorescence indicators, were developed to assess the feasibility of predicting tomato yield under TYLCV infection conditions. The findings of this study are expected to improve the quantitative understanding of productivity loss caused by TYLCV infection and provide a scientific basis for data-driven management strategies and decision-making support in greenhouse tomato production systems.

2. Materials and Methods

2.1. Plant Materials and Cultivation

Two large-fruited red tomato (Solanum lycopersicum L.) cultivars widely cultivated in Korean greenhouse production systems, ‘Daphnis’ and ‘Pink Star’, were used in this study. Experiments were conducted in a plastic-film greenhouse located in Chuncheon, Korea (37.915° N, 127.766° E), under uniform cultivation and management conditions for two consecutive growing seasons. Seeds were sown on 10 January 2021 and 15 January 2022 in commercial seedling trays (72 cells) filled with horticultural substrate. Twenty-one days after sowing, seedlings with fully expanded true leaves were transplanted into rockwool cubes (Plantop 10 × 10 × 10 cm, Grodan Co., Roermond, The Netherlands) and subsequently transferred to rockwool slabs (Growbag 120 × 12 × 7.5 cm, Grodan Co., Roermond, The Netherlands) on 4 March 2021 and 10 March 2022. Planting density was maintained at 6.8 stems m⁻². Plants were trained using a two-stem system commonly applied in greenhouse tomato cultivation. One lateral shoot below the first flower truss was retained as the second stem, while other lateral shoots were removed. Fruit set was induced from the first flower truss, and lower leaves and unnecessary shoots were periodically removed to maintain canopy balance.

The TYLCV inoculum used in this study was obtained in 2021 from greenhouse-grown tomato plants collected from a commercial farm in Chuncheon, Gangwon Province, Korea. The source plants showed typical TYLCV symptoms, including yellowing and curling of young leaves. The presence of TYLCV in the source plants was confirmed by PCR using TYLCV-specific primers. TYLCV inoculation was conducted after the experimental plants were fully established. To minimize growth inhibition caused by early infection, mechanical inoculation was performed on plants that had developed at least the third flower truss. Sap extracted from PCR-confirmed TYLCV-infected leaves was suspended in phosphate buffer and applied to the leaf surfaces, whereas control plants received phosphate buffer only. After inoculation, the experimental plants were monitored for symptom development, and TYLCV infection was re-confirmed by PCR using the same TYLCV-specific primers. Inoculated plants showing positive PCR results exhibited leaf yellowing and curling symptoms consistent with TYLCV infection. Sequencing-based validation of the viral isolate was not performed in this study. To visually document the effects of TYLCV infection on plant morphology, representative images of both healthy and infected plants were captured under greenhouse conditions. Photographs were taken at comparable growth stages to ensure consistency in developmental status. Both canopy structure and shoot apex morphology were recorded to illustrate differences in vegetative growth and symptom expression between treatments. These visual observations are presented in Figure 1.

2.2. Environmental Data Collection

Environmental conditions were monitored using an integrated greenhouse control system (MAXIMIZER 4.2.0 build 4771, Priva B.V., The Netherlands). Air temperature, relative humidity, solar radiation, and CO₂ concentration inside the greenhouse were automatically recorded at 5 min intervals throughout the cultivation period, together with outside temperature and humidity. For environmental analysis, the cultivation period was divided into a high-temperature season (July–August) and a low-temperature season (September–November). Environmental data were averaged daily and summarized into monthly and seasonal means. Annual environmental variation is presented in Figure 2. Greenhouse climate was automatically regulated using heating, ventilation, air circulation, and shading or thermal screens according to a standard management program. Daytime CO₂ concentration was maintained at 400–500 ppm. Irrigation and fertilization were supplied through a non-recirculating fertigation system using a modified Hoagland nutrient solution containing balanced macro- and micronutrients for tomato cultivation. The electrical conductivity (EC) of the nutrient solution was maintained between 2.1 and 2.8 dS m⁻¹. Substrate moisture content was maintained at 55–65% and continuously monitored using sensors inserted into the rockwool medium.

2.3. Growth and Yield Measurements

Growth measurements were conducted weekly during the high-temperature (1 July–31 August) and low-temperature (1 September–31 October) seasons. Ten stems per treatment (TYLCV-infected and healthy control) were selected for each cultivar, resulting in a total of 40 stems that were repeatedly measured throughout the experiment. Measured parameters included plant height increment, truss height, stem diameter, leaf length, leaf width, leaf number, leaf area, number of flowering trusses, number of fruit-setting trusses, and yield. Plant height was defined as the vertical distance from the rockwool slab surface to the apical meristem, and growth increment was calculated from successive measurements. Stem diameter was measured below the first flowering truss, while leaf length and width were measured from a representative leaf immediately below the flowering truss. The number of flowering trusses was defined as trusses with at least one fully opened flower or fruits smaller than 2 cm in diameter, whereas fruit-setting trusses contained fruits larger than 2 cm.

Leaf area index (LAI) was estimated non-destructively using the empirical equation proposed by Jang et al. (2021) [15]:

LAI = Leaf length (cm) × Leaf width (cm) × 0.6 × SD × (L1 + 0.7L2)

(1)

where SD represents planting density (stems m⁻²), L1 represents leaf number on the main stem, and L2 represents leaf number on the lateral stem. Leaf length and width were measured from representative leaves at the same developmental position in all treatments to ensure consistency in comparative evaluation. In this study, the estimated LAI values were used primarily as a relative indicator of canopy development among treatments rather than as an absolute measure of leaf area. Because TYLCV infection may induce leaf yellowing and curling, which could influence the accuracy of equation-based estimation, the LAI results should be interpreted with caution under infected conditions.

Fruits were harvested when surface coloration exceeded 80%. The number of harvested fruits and total fruit weight were recorded at each harvest. Total yield per plant was calculated as cumulative fruit weight during the cultivation period, and average fruit weight was obtained by dividing total yield by fruit number. Yield values were expressed as yield per stem for statistical comparison.

To quantify yield loss caused by TYLCV infection, relative reduction (%) was calculated as

Relative reduction (%) = (Healthy mean − Virus mean)/Healthy mean × 100

(2)

where Healthy mean and Virus mean represent the average values of healthy and TYLCV-infected plants under the same cultivar and seasonal conditions.

2.4. Physiological Measurements

Physiological responses were evaluated using chlorophyll fluorescence and photosynthetic gas exchange measurements. Measurements were conducted on leaves at the same developmental stage following TYLCV inoculation. One representative leaf per plant (the second fully expanded leaf) was selected for measurement to ensure consistency in the developmental stage. Chlorophyll fluorescence was measured on the second fully expanded leaf using a portable fluorometer (Junior-PAM, Heinz Walz GmbH, Germany). Leaves were dark-adapted for 20 min before measurement. A saturating pulse of blue light (445 nm, 2800 µmol m⁻² s⁻¹) was applied to record minimum fluorescence (Fo), maximum fluorescence (Fm), and variable fluorescence (Fv = Fm − Fo). The maximum quantum efficiency of photosystem II (Fv/Fm) was calculated as

Fv/Fm = (Fm − Fo)/Fm

(3)

To evaluate electron transport characteristics in photosystem II, chlorophyll fluorescence OJIP transient analysis was also conducted.

Photosynthetic gas exchange was measured using a portable photosynthesis system (LI-6800, LI-COR Biosciences, Lincoln, NE, USA). Measurements were conducted between 07:00 and 09:00 to minimize diurnal variation. Fully expanded leaves without visible damage were placed in the chamber, and data were recorded after gas exchange stabilized.

CO₂ response curves (A–Ci curves) were obtained by stepwise adjustment of chamber CO₂ concentration. Net photosynthetic rate (A; µmol m⁻² s⁻¹) and intercellular CO₂ concentration (Ci; Pa) were recorded at each step to construct A–Ci curves. Intercellular CO₂ concentration at 300 ppm (Ci300) and maximum photosynthetic rate (Amax) were used as key physiological indicators. Relationships among yield and physiological or growth traits were analyzed using GGE and GT biplot analyses based on principal component analysis (PCA). Trait relationships and genotype responses were visualized using the first two principal components (PC1 and PC2). All analyses were performed using R statistical software (R version 4.3.1).

2.5. Yield Prediction Model

To explain and predict yield variation under TYLCV infection, a multiple linear regression model was developed using physiological and growth traits. Total yield per stem (g) was used as the dependent variable. Pearson correlation analysis was first conducted to identify traits significantly associated with yield. Variables showing significant correlations (Amax, Fv/Fm, Ci300, and truss height) were selected as candidate predictors. Multicollinearity among variables was evaluated using variance inflation factors (VIFs), and variables with VIF values exceeding 10 were excluded.

Stepwise regression was applied to determine the final predictors. In the final model, Fv/Fm and Ci300 were selected as key explanatory variables. TYLCV infection status was included as a dummy variable (healthy = 0, infected = 1). Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). RMSE reflects the square root of the average squared prediction errors and is more sensitive to large deviations, whereas MAE represents the average absolute difference between observed and predicted values and provides a more robust measure of overall prediction accuracy. Regression assumptions were assessed using residual distribution and normal Q–Q plots. All analyses were performed using R software.

2.6. Statistical Analysis

Growth, physiological, and yield data were analyzed using two-way analysis of variance (ANOVA) with cultivar and seasonal environment as factors. Mean comparisons were performed using Tukey’s HSD test at p < 0.05, and effect sizes were evaluated using partial η². Relationships among traits were examined using Pearson correlation analysis. GGE and GT biplot analyses were conducted to visualize genotype–environment interactions and trait relationships. All statistical analyses and graphical visualizations were performed using R statistical software (R version 4.3.1).

3. Results

3.1. Seasonal Greenhouse Environmental Conditions

Greenhouse environmental conditions during the tomato cultivation periods in 2020 and 2021 showed clear seasonal variation between the defined high-temperature season (weeks 27–35) and low-temperature season (weeks 36–44) (Figure 2). Weekly mean air temperature remained high during the high-temperature season and gradually declined after the transition to the low-temperature season (Figure 2A,B). During the mid–high-temperature period, both the average temperature and the range between weekly minimum and maximum temperatures increased. In contrast, temperature levels and variability decreased during the low-temperature season. Similar patterns were observed in both years, indicating that seasonal effects rather than interannual differences primarily influenced greenhouse temperature dynamics. Relative humidity showed a gradual increase as the cultivation period progressed (Figure 2C,D). During the high-temperature season, relative humidity exhibited relatively large weekly fluctuations, whereas it remained higher and more stable during the low-temperature season. Cumulative solar radiation also exhibited a clear seasonal pattern (Figure 2E,F). Radiation levels remained high during the high-temperature season and decreased sharply after the transition to the low-temperature season. Lower radiation levels were maintained throughout the low-temperature period. Overall, the cultivation period clearly formed two contrasting seasonal greenhouse environments characterized by simultaneous changes in temperature, relative humidity, and solar radiation (Figure 2A–F). These environmental contrasts provided the background conditions for interpreting seasonal differences in plant growth, physiological responses, and yield under TYLCV infection.

3.2. Growth and Yield Responses to TYLCV Infection

TYLCV infection consistently affected tomato growth and yield-related traits, although the magnitude of these responses varied depending on season and cultivar (

Table 1. Effects of TYLCV infection on vegetative growth traits of two tomato cultivars under high- and low-temperature seasons in greenhouse cultivation.

Season	Cultivar	Virus	Weekly Growth (cm)	Flower Cluster Height (cm)	Stem Diameter (mm)	Leaf Length (cm)	Leaf Width (cm)	LAI (cm2)	Number of Leaf
High	Daphnis	Healthy	27.28 ± 0.29 a	19.35 ± 0.22 abc	11.23 ± 0.20 a	51.42 ± 0.38 a	47.71 ± 0.61 a	102,776.32 ± 3381.24 a	10.04 ± 0.19 a
High	Daphnis	TYLCV	25.02 ± 0.26 b	20.25 ± 0.23 a	10.33 ± 0.19 bcd	46.08 ± 0.29 c	42.56 ± 0.54 d	71,042.14 ± 2233.42 c	8.72 ± 0.16 b
High	Pink star	Healthy	24.85 ± 0.27 b	17.64 ± 0.20 de	10.22 ± 0.19 bcd	48.07 ± 0.36 b	44.60 ± 0.57 bc	80,857.61 ± 2726.35 b	9.05 ± 0.18 b
High	Pink star	TYLCV	22.79 ± 0.25 c	18.47 ± 0.21 cd	9.46 ± 0.16 d	43.15 ± 0.26 d	39.79 ± 0.50 e	55,512.11 ± 1731.92 ef	7.80 ± 0.14 c
Low	Daphnis	Healthy	17.68 ± 0.39 d	18.69 ± 0.31 bcd	11.08 ± 0.23 ab	50.74 ± 0.28 a	46.33 ± 0.37 ab	84,186.97 ± 1763.76 b	8.75 ± 0.16 b
Low	Daphnis	TYLCV	16.42 ± 0.36 de	19.68 ± 0.32 ab	10.37 ± 0.22 abc	46.18 ± 0.26 c	42.05 ± 0.33 d	61,514.16 ± 1380.59 de	7.70 ± 0.15 c
Low	Pink star	Healthy	16.20 ± 0.36 ef	17.15 ± 0.28 e	10.15 ± 0.21 cd	47.83 ± 0.26 b	43.55 ± 0.34 cd	66,892.33 ± 1431.11 cd	7.84 ± 0.15 c
Low	Pink star	TYLCV	14.97 ± 0.33 f	18.02 ± 0.29 de	9.46 ± 0.20 d	43.53 ± 0.24 d	39.63 ± 0.31 e	49,066.22 ± 1066.64 f	6.91 ± 0.13 d

Values are presented as mean ± standard error (SE) based on 10 biological replicates (n = 10). Different lowercase letters within the same column indicate significant differences among treatments according to Tukey’s HSD test at p < 0.05. High season represents July–August, and low season represents September–October during the cultivation period. SE indicates the variability of the mean among biological replicates. LAI, leaf area index; TYLCV, tomato yellow leaf curl virus.

,

Table 2. Effects of tomato yellow leaf curl virus (TYLCV) infection on reproductive and yield traits of two tomato cultivars under high- and low-temperature seasons in greenhouse cultivation.

Season	Cultivar	Virus	Bloom Flower Count	Flower Cluster Fruit Count	Yield (g)	Average Fruit Weight (g)	Total Yield (g)
High	Daphnis	Healthy	1.41 ± 0.09 a	1.62 ± 0.08 a	837.59 ± 9.30 c	167.18 ± 1.98 c	3765.29 ± 39.59 c
High	Daphnis	TYLCV	0.96 ± 0.04 b	1.00 ± 0.04 b	676.20 ± 9.63 d	133.24 ± 1.93 e	3115.94 ± 39.43 d
High	Pink star	Healthy	1.54 ± 0.09 a	1.54 ± 0.10 a	1075.43 ± 8.88 a	215.09 ± 1.78 a	4837.65 ± 37.83 a
High	Pink star	TYLCV	0.89 ± 0.08 b	0.95 ± 0.08 b	855.47 ± 7.63 bc	171.09 ± 1.53 c	3954.34 ± 32.87 b
Low	Daphnis	Healthy	1.41 ± 0.09 b	1.60 ± 0.07 b	840.09 ± 9.44 c	157.52 ± 1.77 d	3720.70 ± 38.28 c
Low	Daphnis	TYLCV	0.96 ± 0.08 a	1.00 ± 0.07 a	709.84 ± 9.83 d	133.09 ± 1.84 e	3183.44 ± 39.12 d
Low	Pink star	Healthy	1.54 ± 0.09 b	1.54 ± 0.10 b	1071.62 ± 9.48 a	200.93 ± 1.78 b	4748.20 ± 42.23 a
Low	Pink star	TYLCV	0.96 ± 0.04 a	1.00 ± 0.04 a	895.44 ± 8.72 b	167.90 ± 1.64 c	4021.84 ± 38.82 b

Values are presented as mean ± standard error (SE) based on 10 biological replicates (n = 10). Different lowercase letters within the same column indicate significant differences among treatments according to Tukey’s HSD test at p < 0.05. High season represents July–August and low season represents September–October during the cultivation period. SE indicates the variability of the mean among biological replicates. Yield represents fruit weight per harvest event (g), whereas total yield represents cumulative fruit yield per stem (g). TYLCV, tomato yellow leaf curl virus.

and

Table 3. Relative reduction (%) in reproductive and yield traits of two tomato cultivars following TYLCV infection under high- and low-temperature seasons in greenhouse cultivation.

Season	Cultivar	Bloom Flower Count	Flower Cluster Fruit Count	Yield (g)	Average Fruit Weight (g)	Total Yield (g)
High	Daphnis	31.86	38.46	19.27	19.10	17.25
High	Pink star	42.28	38.21	20.45	20.45	18.26
Low	Daphnis	−46.75	−60.00	15.50	15.50	14.44
Low	Pink star	−59.74	−53.75	16.44	16.44	15.30

Values represent the relative reduction (%) calculated from treatment means for each trait under the same cultivar and seasonal condition. Relative reduction (%) was calculated as (Healthy mean − Virus mean)/Healthy mean × 100. Because these values were derived from group means, measures of variation were not separately presented in this table. Positive values indicate a decrease in the trait due to TYLCV infection, whereas negative values indicate an increase relative to the healthy control. High season represents July–August and low season represents September–October during the cultivation period. TYLCV, tomato yellow leaf curl virus.

, Tables S1 and S2). Infected plants generally showed reduced growth compared with healthy controls (Table 1). Weekly growth increment, leaf length, leaf width, leaf area index (LAI), and leaf number were noticeably lower in TYLCV-infected plants. This pattern was observed in both cultivars, with the largest reductions occurring in leaf-related traits, particularly LAI, leaf length, and leaf width. In contrast, flower cluster height showed relatively small or inconsistent differences between infection treatments. Two-way ANOVA indicated significant effects of season and cultivar for most growth traits (Table S1). Seasonal effects were particularly strong for weekly growth increment and plant height (partial η² = 0.95 and 0.91, respectively), suggesting that growth responses were strongly influenced by environmental conditions. In contrast, cultivar effects were more pronounced for leaf length, leaf width, and LAI (partial η² = 0.79–0.91), indicating that genetic differences between cultivars contributed substantially to variation in leaf morphology. The season × cultivar interaction was generally non-significant or relatively small, suggesting that TYLCV-induced growth reduction occurred consistently across seasonal environments.

Yield-related traits also showed clear reductions in responses to TYLCV infection (Table 2). Infected plants exhibited lower values for bloom flower count, flower cluster fruit count, average fruit weight, and total yield compared with healthy plants. In both cultivars, TYLCV infection affected fruit formation and fruit enlargement processes, ultimately leading to reduced total yield. Among the yield traits, total yield showed the largest difference between cultivars, with ‘Pink Star’ consistently maintaining higher yield levels than ‘Daphnis’. Two-way ANOVA further confirmed a strong cultivar effect on yield-related traits (Table S2). In particular, cultivar effects for yield, average fruit weight, and total yield showed very large effect sizes (partial η² > 0.96), indicating that cultivar was the dominant factor influencing variation in yield traits. In contrast, seasonal effects were limited to a few traits. However, relatively large season × cultivar interaction effects were observed for bloom flower count and flower cluster fruit count (partial η² = 0.56–0.64), suggesting that reproductive responses differed between cultivars depending on seasonal environmental conditions.

To quantitatively evaluate yield reduction caused by TYLCV infection, the relative reduction (%) of yield-related traits was calculated (Table 3). During the high-temperature season, both cultivars exhibited relatively large reductions in yield-related traits. In particular, bloom flower count and flower cluster fruit count decreased by approximately 31–42%, while total yield declined by about 17–18%. In contrast, the magnitude of yield reduction was generally smaller during the low-temperature season. In some traits, values in TYLCV-infected plants were comparable to or slightly higher than those of healthy controls. These results indicate that although TYLCV infection negatively affects reproductive traits and yield formation in tomato, the extent of its impact varies depending on environmental conditions and cultivar characteristics. Overall, TYLCV infection reduced tomato yield primarily by affecting leaf growth and reproductive development. However, the magnitude of these responses appeared to be modulated by the interaction between cultivar characteristics and seasonal environmental conditions.

3.3. Physiological Responses

TYLCV infection induced consistent physiological changes in the photosynthetic system of tomato leaves (Figure 2 and Figure 3;

Table 4. Effects of tomato yellow leaf curl virus (TYLCV) infection on the maximum quantum efficiency of photosystem II (Fv/Fm) in two tomato cultivars under high- and low-temperature seasons.

Season	Cultivar	Healthy (Fv/Fm)	TYLCV (Fv/Fm)	Reduction (%)
High	Daphnis	0.812 ± 0.002 a	0.719 ± 0.002 c	11.48
High	Pink star	0.812 ± 0.002 a	0.721 ± 0.002 c	11.17
Low	Daphnis	0.813 ± 0.002 a	0.749 ± 0.005 b	7.89
Low	Pink star	0.813 ± 0.002 a	0.752 ± 0.005 b	7.48
ANOVA (Fv/Fm) z
Season (A)			***
Cultivar (B)			NS
Virus (C)			***
A × B			NS
A × C			***
B × C			NS
A × B × C			NS

Values are presented as mean ± standard error (SE). Different lowercase letters within the same column indicate significant differences among treatments according to Tukey’s HSD test at p < 0.05. ^z *, **, and *** indicate significance at p < 0.05, p < 0.01, and p < 0.001, respectively. NS indicates not significant.

). Chlorophyll fluorescence OJIP transient analysis showed that infected plants tended to exhibit a lower increase in fluorescence intensity along the O–J–I–P phases compared with healthy controls (Figure 3). This pattern indicates reduced electron transport efficiency in photosystem II. Similar responses were observed in both cultivars, and fluorescence intensity in infected plants remained lower than that of healthy controls at the P phase, where fluorescence typically reaches saturation. These results suggest that TYLCV infection may impose structural or functional constraints on the photochemical processes of photosystem II.

These changes in fluorescence responses were also reflected in the reduction in the maximum quantum yield of photosystem II (Fv/Fm) (Table 4). In healthy plants, Fv/Fm values of both cultivars were approximately 0.81, indicating stable photosynthetic efficiency. In contrast, TYLCV-infected plants showed reductions of approximately 11% during the high-temperature season and 7–8% during the low-temperature season. Two-way ANOVA revealed significant effects of both season and virus factors (p < 0.001), with a significant season × virus interaction, indicating that the influence of TYLCV infection on photochemical efficiency varied depending on environmental conditions. In contrast, the cultivar effect was not statistically significant, suggesting that the basic photosystem II response patterns were similar between the two cultivars.

Photosynthetic carbon assimilation was also clearly reduced by TYLCV infection (Figure 4). Analysis of A–Ci response curves showed that infected plants exhibited lower net photosynthetic rates (A) than healthy plants at similar intercellular CO₂ concentrations, and the difference became more pronounced as Ci increased. Under high-temperature conditions, infected plants showed a markedly lower photosynthetic saturation level, indicating that TYLCV infection limited the upper capacity of carbon assimilation. This response was consistently observed in both cultivars. Although a similar trend was also detected under low-temperature conditions, the magnitude of reduction was relatively smaller than that observed during the high-temperature season.

3.4. G × E and GGE Biplot Analysis

GGE biplot and GT biplot analyses were performed to comprehensively evaluate the relationships between total yield and major physiological and growth traits (Figure 4, Figure 5 and Figure 6, Figures S1 and S2).

The GGE biplot based on total yield showed that the first principal component (PC1) explained the majority of the total variation and primarily represented differences in yield performance among genotypes (Figure 5). In contrast, the second principal component (PC2) explained a relatively smaller proportion of variation and was interpreted as reflecting differences in responses to environmental conditions. The two cultivars were clearly separated along the PC1 axis, indicating that yield variation was mainly explained by genotype effects. ‘Pink Star’ was positioned in the positive direction of PC1, indicating generally higher yield potential, whereas ‘Daphnis’ was located in a region associated with relatively lower productivity. TYLCV-infected plants showed positional shifts compared with healthy plants of the same cultivar, suggesting that virus infection influenced genotype-specific yield performance patterns. Environmental vectors representing seasonal conditions were arranged along the positive PC1 direction, indicating that both high- and low-temperature seasons contributed to yield variation but did not substantially alter the relative ranking of genotypes.

The polygon view of the GGE biplot further illustrated genotype-specific yield responses under different seasonal environments (Figure 6). Genotypes located at the polygon vertices represent those with relatively high yield potential under specific environmental conditions. In this study, environmental conditions were distributed within a relatively narrow region of the biplot space, indicating that environmental variation had a relatively smaller influence on yield differences compared with genotype effects under the tested greenhouse conditions. This suggests that yield variation in this experiment was largely driven by genotype rather than environmental factors.

To examine relationships among traits associated with yield formation, GT biplot analysis was conducted using total yield together with key physiological and morphological traits (Figure 7). The GT biplot showed that total yield was positioned in a similar direction to average fruit weight and leaf area index (LAI), indicating a strong association with canopy development and vegetative growth traits. These results suggest that even under TYLCV infection, yield formation remained closely linked to plant growth and canopy structure. In contrast, TYLCV-infected plants tended to occupy different regions of the biplot compared with healthy plants, indicating that virus infection may alter the relationships among growth and physiological traits.

Supplementary GT biplot analyses showed similar patterns. In the growth trait–based GT biplot (Figure S1), plant height, leaf area index, and related growth traits clustered in the same direction, indicating strong associations among vegetative growth responses. In the photosynthetic trait-based GT biplot (Figure S2), photosynthetic capacity indicators such as A300, A1600, and Amax were aligned in the same direction, suggesting close relationships among traits associated with carbon assimilation capacity. In contrast, Fv/Fm was located near the origin relative to other photosynthetic traits, indicating that the maximum photochemical efficiency of photosystem II contributed relatively less to the major variation explained in this study.

Overall, the GGE and GT biplot analyses indicated that the variation in tomato yield observed in this study was primarily explained by genotype effects rather than seasonal environmental conditions. In addition, yield formation was closely associated with plant growth characteristics and photosynthesis-related traits.

3.5. Physiological Trait-Based Yield Prediction Model

The relationships between total yield and major physiological and growth traits were evaluated to identify key predictors of tomato yield (

Table 5. Correlation of total yield with key physiological and growth traits in tomato plants under TYLCV infection.

Trait	Pearson r	p Value
Amax	0.57484	0.00000
Fv/Fm	0.52859	0.00000
Ci300	0.42723	0.00000
Flower cluster height (cm)	−0.27897	0.00035

Traits showing statistically significant correlations with total yield (p < 0.05) are presented. However, correlation strength should be interpreted based on the magnitude of Pearson’s r.

; Figure 8). Total yield showed the strongest positive correlation with maximum net photosynthetic rate (Amax) (r = 0.5748, p < 0.001). The maximum quantum efficiency of photosystem II (Fv/Fm) was also positively correlated with yield (r = 0.5286, p < 0.001). In addition, intercellular CO₂ concentration at 300 ppm (Ci300) derived from the A–Ci curve showed a significant positive relationship with yield (r = 0.4272, p < 0.001). In contrast, flower cluster height showed only a weak negative correlation with yield (r = −0.2789, p = 0.00035), and this relationship should therefore be interpreted with caution. These relationships were illustrated in the scatterplots shown in Figure 8, where increased photosynthetic capacity was associated with higher yield, whereas the relationship between flower cluster height and yield was weak and should not be overinterpreted.

Based on these trait relationships, a multiple regression model was developed to predict tomato yield using physiological indicators (

Table 6. Multiple regression coefficients for the physiological trait-based model predicting total yield in tomato plants under TYLCV infection conditions.

Variable	Coefficient	Std Error	t Value	p Value	VIF
Intercept	5385.26	3091.97	1.74	0.084	5749.55
Fv/Fm	2247.36	2644.99	0.85	0.397	7.55
Ci300	−31.70	36.95	−0.86	0.392	3.44
Virus (dummy)	−628.57	200.14	−3.14	0.002	6.02

Values represent regression coefficients from a multiple linear regression model predicting total yield per stem (g). Std error, t value, and p value indicate the statistical significance of each predictor. Virus infection was included as a dummy variable (healthy = 0, TYLCV-infected = 1), and VIF values were calculated to assess multicollinearity among predictors. TYLCV, tomato yellow leaf curl virus.

). Among the initial candidate variables, variance inflation factor analysis was used to remove multicollinearity, and Fv/Fm, Ci300, and virus infection status (dummy variable) were retained in the final model. Regression analysis indicated that virus infection status was a significant predictor of yield (p = 0.002), with infected plants showing lower total yield compared with healthy plants. In contrast, Fv/Fm and Ci300 were not statistically significant predictors but showed positive and negative associations with yield, respectively, suggesting that physiological status may contribute to yield variation to some extent.

The relative contribution of each predictor was further evaluated using standardized regression coefficients (standardized β) (Figure 9). Virus infection status showed the largest absolute β value, indicating that disease stress was the dominant factor influencing yield variation. In contrast, Fv/Fm showed a positive relationship with yield, whereas Ci300 and flower cluster height exhibited relatively small negative effects. These results suggest that viral stress acts as a major limiting factor for yield formation, while photosynthetic efficiency and carbon assimilation traits contribute to the physiological basis of yield variability.

Model performance evaluation showed moderate explanatory power between observed and predicted yield values (R² = 0.367), with a root mean square error (RMSE) of 492.46 g and a mean absolute error (MAE) of 439.49 g (Figure 10). The relatively higher RMSE compared with MAE suggests that the model was influenced by a small number of observations with relatively large prediction errors, indicating the presence of potential outliers or variability under specific conditions. These results indicate that the model explains part of the variation in tomato yield, while additional factors such as environmental conditions and cultivar characteristics may also influence yield formation.

Residual diagnostics were conducted to verify the statistical assumptions of the regression model (Figure 11). The residuals versus fitted plot showed no clear pattern or systematic increase in variance, suggesting that the assumption of homoscedasticity was largely satisfied. In addition, the normal Q–Q plot showed that most residuals were distributed close to the reference line, indicating that the residuals approximately followed a normal distribution. These results suggest that the developed regression model generally satisfied the statistical assumptions required for yield prediction.

4. Discussion

This study investigated the effects of tomato yellow leaf curl virus (TYLCV) infection on plant growth, physiological responses, and yield formation in greenhouse-grown tomatoes and evaluated the potential of physiological traits for predicting yield under virus infection conditions. The results demonstrated that TYLCV infection influences not only visible disease symptoms but also key physiological and structural processes, including photosynthetic performance, canopy development, and reproductive formation, ultimately leading to yield reduction [1,16]. By integrating physiological indicators related to photosynthesis with plant growth traits, this study provides a more comprehensive understanding of the physiological mechanisms through which viral infection affects tomato productivity.

TYLCV infection significantly altered the photosynthetic performance of tomato leaves. In the chlorophyll fluorescence OJIP transient analysis, infected plants exhibited reduced fluorescence increases during the J–I–P phases, indicating decreased electron transport efficiency in photosystem II (PSII) [17,18]. Changes in the OJIP fluorescence curve are commonly used as indicators of photochemical efficiency and electron transport dynamics in the photosynthetic apparatus. The reduction in Fv/Fm further confirmed that TYLCV infection decreased the maximum photochemical efficiency of PSII [19,20]. Similar responses have been reported in virus-infected plants, where viral infection can induce structural changes in chloroplasts or reduce chlorophyll content, both of which are closely associated with declines in photosynthetic efficiency. Therefore, the fluorescence responses observed in this study suggest that TYLCV infection may impose structural or functional constraints on the photosynthetic apparatus.

Alterations in the photosynthetic system were also reflected in reduced carbon assimilation capacity. The A–Ci response curve analysis showed that TYLCV-infected plants exhibited lower net photosynthetic rates under similar intercellular CO₂ concentrations, particularly at higher Ci levels. This pattern suggests that TYLCV infection may influence not only stomatal regulation but also the biochemical processes of photosynthesis. Reduced photosynthetic rates are generally associated with factors such as decreased Rubisco activity, limitations in electron transport, or impaired chloroplast function [21,22]. Viral infection may simultaneously affect these processes by reducing stomatal conductance and altering chloroplast structure and photosynthetic enzyme activity, thereby resulting in both stomatal and non-stomatal limitations in photosynthesis.

Reduced photosynthetic performance may also influence plant growth and canopy structure. In the present study, TYLCV infection decreased leaf length, leaf width, and leaf area index (LAI), which may be associated with reduced assimilate production. In greenhouse tomato production systems, canopy structure and LAI are key traits determining the crop’s capacity to intercept light. A reduction in LAI decreases light interception per unit ground area and may consequently limit the carbon supply required for fruit enlargement [9]. Consistent with this interpretation, GT biplot analysis showed that total yield was positioned in a similar direction to LAI and average fruit weight, indicating a close relationship between canopy development and fruit production [14]. These results suggest that TYLCV infection affects yield formation through both reduced photosynthetic capacity and altered canopy structure [10]. In addition, because LAI was estimated using an empirical equation based on leaf length and width, the accuracy of LAI estimation may have been affected by TYLCV-induced leaf deformation, such as yellowing and curling. Therefore, the LAI values in this study should be interpreted primarily as a relative indicator of canopy development rather than an absolute measure of leaf area.

The relative contributions of genotype and environmental conditions to yield variation were also evaluated. The GGE biplot analysis indicated that yield variation was more strongly explained by genotype effects than by seasonal environmental conditions. This finding suggests that productivity differences among cultivars may occur even under similar environmental conditions [5]. In this study, the cultivar ‘Pink Star’ consistently showed higher yield potential than ‘Daphnis’, and its productivity remained relatively stable even under TYLCV infection conditions. These results imply that cultivar-specific characteristics such as growth vigor, photosynthetic capacity, or carbon allocation patterns may contribute to yield stability [23]. Although many previous TYLCV studies have primarily focused on resistance genes, the present results suggest that physiological traits may also play an important role in explaining cultivar-specific productivity differences [24].

Seasonal environmental conditions also influenced the magnitude of TYLCV effects. TYLCV replication and symptom expression are generally known to be enhanced under high-temperature conditions. In the present study, yield reduction was more pronounced during the high-temperature season, indicating that environmental stress may amplify the negative effects of viral infection. Under high-temperature conditions, plants often experience increased respiration and reduced photosynthetic efficiency. When viral infection occurs simultaneously, the plant carbon balance may be further disrupted, leading to greater reductions in growth and productivity [6,25]. Therefore, the interaction between viral infection and temperature stress may intensify physiological constraints on crop performance [26].

Another objective of this study was to evaluate the potential of physiological traits for predicting tomato yield. Correlation analysis showed that maximum net photosynthetic rate (Amax) and Fv/Fm were positively associated with total yield, highlighting the importance of photosynthetic performance in determining crop productivity [11]. In the multiple regression analysis, virus infection status was identified as the most influential predictor of yield variation, indicating that disease stress acts as a major limiting factor for tomato production [27]. Although the explanatory power of the regression model was moderate (R² = 0.367), the results demonstrate that physiological trait-based models can provide a useful framework for predicting yield under virus infection conditions. Similar modeling approaches have also been reported in tomato under non-infected conditions, including process-based greenhouse growth models and fruit growth prediction models, which have been used to improve model localization, cultivar-specific calibration, and predictive precision under controlled environments [28,29].

However, the developed model did not fully explain the observed variation in tomato yield, indicating several limitations of the present study. First, although TYLCV infection was confirmed by PCR in both source and inoculated plants, sequencing-based validation of the viral isolate was not performed. In addition, the possibility of mixed infection with other mechanically transmissible viruses cannot be completely excluded. These factors may have introduced uncertainty in the interpretation of virus-specific effects on physiological responses and yield formation. Second, the regression model showed only moderate explanatory power (R² = 0.367), suggesting that yield formation in greenhouse tomato production is influenced by multiple interacting factors beyond the physiological traits considered in this study. Environmental conditions, crop management practices, plant architectural traits, and genetic variability among cultivars may all contribute to yield variation and were not fully captured in the current model framework. Third, the estimation of leaf area index (LAI) was based on an empirical equation using leaf length and width, which may be affected by TYLCV-induced leaf deformation such as yellowing and curling. Therefore, LAI values should be interpreted primarily as a relative indicator of canopy development rather than an absolute measure of leaf area under virus infection conditions.

Future research should aim to address these limitations by incorporating more comprehensive datasets and validation approaches. In particular, sequencing-based virus identification and screening for potential co-infections would improve the reliability of infection characterization. In addition, integrating environmental variables, canopy structural parameters, and genotype-specific traits into modeling frameworks may enhance predictive accuracy. With the advancement of smart farming and precision agriculture technologies, real-time monitoring of plant physiological responses could further support the development of robust, data-driven yield prediction models under both biotic and abiotic stress conditions.

Overall, this study demonstrates that TYLCV infection reduces tomato productivity by simultaneously affecting photosynthetic performance, canopy development, and reproductive processes. In addition, the integration of physiological traits provided a quantitative framework for understanding the mechanisms underlying yield reduction and highlighted the potential of physiological trait-based approaches for predicting crop productivity under viral infection conditions. These findings may contribute to improving disease management strategies and productivity forecasting in greenhouse tomato production systems.

5. Conclusions

This study demonstrated that TYLCV infection reduces tomato productivity by impairing photosynthetic efficiency and canopy development, which subsequently affects plant growth and fruit formation. These physiological changes ultimately contribute to yield reduction in greenhouse-grown tomatoes. In addition, the regression model developed using photosynthesis-related physiological traits and virus infection status suggested that yield variation under TYLCV infection conditions can be quantitatively explained to a certain extent. These findings provide a useful foundation for evaluating crop productivity under viral infection conditions and highlight the potential of physiological trait-based approaches for data-driven yield prediction in greenhouse tomato production systems.