Intraspecific variation in morphological and defensive traits of guava genotypes and their implications for herbivory

Intraspecific genotypic variation arising from artificial selection can shift how plants allocate resources, often unbalancing traits related to growth, reproduction, and defense. Such trade-offs may increase susceptibility to herbivory, particularly in genotypes subjected to strong selection. While perennial plants often retain ancestral traits, the effects of selective breeding on their morphological and chemical characteristics—and their consequences for plant–herbivore interactions—remain poorly understood. Here, we examined how intraspecific genotypic variation shaped by artificial selection influences morphological and biochemical traits in guava (Psidium guajava), and how these traits affect herbivory. We evaluated seven guava genotypes, two wild, two landraces, and three cultivated to capture genetic variation using material from a germplasm bank in Zacatecas, Mexico. We measured tree, leaf, and fruit morphology; total soluble solids (°Brix); water content; and total phenolic compounds in leaves and fruits. Folivory and frugivory were also quantified. Most traits varied significantly among genotypes. Cultivated guavas generally had larger leaves, greater fruit biomass, and heavier seeds, but lower concentrations of chemical defenses in both leaves and fruits. However, no trade-offs were detected between morphological and defensive traits. Foliar herbivory varied among genotypes but was not associated with the origin of the plant. Moreover, herbivory showed no correlation with any morphological or biochemical traits. Although artificial selection appears to have reduced certain defensive traits in some guava genotypes, the extent to which it has shaped trait expression and herbivore resistance in wild versus cultivated populations remains unclear.

Intraspecific genotypic variation in resistance to insect herbivory is a critical aspect of plant evolution and adaptation (Schoonhoven et al., 2005; Futuyma and Agrawal, 2009). As sessile organisms, plants have evolved a wide array of defense mechanisms to cope with herbivore pressure. These defenses range from direct physical and chemical traits—such as leaf toughness, trichomes, spines, and toxic or repellent compounds—to indirect defenses like volatile organic compounds (VOCs) and extrafloral nectaries that attract natural enemies of insect herbivores (Agrawal, 2007; Fernández et al., 2021). However, selective breeding can introduce both ecological and physiological trade-offs.

The resource-allocation trade-off hypothesis posits that increased investment in traits such as yield can reduce the resources available for defense (Saeki et al., 2014; Whitehead et al., 2017). For instance, artificial selection can cause phenotypic divergence between incipiently selected and cultivated genotypes due to changes in agricultural practices. Highly selected plants are generally expected to be more susceptible to herbivores and pathogens than their wild relatives, owing to a loss of resistance traits and a shift toward enhanced vegetative growth, yield, or other agronomic characteristics (Rosenthal and Dirzo, 1997; Chen et al., 2015; Stenberg et al., 2015; Whitehead et al., 2017). While reduced defense levels in intensively selected genotypes often lead to greater herbivore preference and damage (Turcotte et al., 2014; Whitehead et al., 2017; Fernandez et al., 2021), this effect can vary depending on plant species or cultivar, herbivore feeding guild, the plant tissue consumed, and host specificity (Gols et al., 2008; Gaillard et al., 2017; Hernández-Cumplido et al., 2018, 2021, Fernandez et al., 2021). In annual cultivated plants, weaker defenses against herbivores are generally expected because these species prioritize rapid growth and high yield within a single growing season, often at the expense of resistance traits. In contrast, perennial species must withstand long-term herbivore pressure, which selects for stronger and more durable defensive investments to ensure survival and reproduction across multiple seasons (Moles et al., 2011; Turcotte et al., 2017; Züst and Agrawal, 2017; Whitehead et al., 2017).

Guava (Psidium guajava L., Myrtaceae) is a perennial tree whose artificial selection history offers valuable insights into plant defense dynamics in long-lived crop systems. Thought to have originated in the Brazilian Amazon, guava is now widely cultivated in tropical and subtropical regions, with Mexico ranking as the third-largest producer (Arévalo-Marín et al., 2021). Breeding efforts have primarily focused on improving fruit quality and yield, enhancing resistance to diseases, and increasing adaptability to diverse environmental conditions (Padilla-Ramirez et al., 2007, 2010, 2014, 2016). In Mexico, guava production is threatened by several insect pests, most notably the guava weevil (Conotrachelus dimidiatus Champion; Coleoptera: Curculionidae), a specialist frugivore that feeds exclusively on Psidium species (González-Gaona et al., 2002; Salas-Araiza and Romero-Nápoles, 2012, Tafoya, 2013, Vargas-Madríz et al., 2018), and fruit flies of the genus Anastrepha, generalist frugivores with a broad host range (Servicio Nacional de Sanidad, Inocuidad y Calidad Agroalimentaria (SENASICA), 2025).

This study investigates whether intraspecific genotypic variation linked to the artificial selection on guava influences resistance to herbivores through reductions in chemical defenses in leaves and fruits. Conducted in Mexico, the research: (1) characterized morphological traits of leaves and fruits across seven guava genotypes with varying domestication histories; (2) analyzed water content, total soluble solids (°Brix), and phenolic compound concentrations in the leaves and fruits of these genotypes; and (3) assessed whether variation in these morphological and biochemical traits correlates with levels of folivory and frugivory. Based on the resource allocation trade-off hypothesis, we expected more domesticated developed guava genotypes to exhibit lower levels of defense and higher rates of herbivory compared to their less domesticated counterparts (Figure 1). According to the hypothesis, selection for traits associated with yield and fruit quality during domestication may come at the cost of reduced investment in defensive traits, such as leaf toughness and phenolic compounds.

Study site

The research was conducted at the experimental campus Los Cañones of the Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), located in the municipality of Huanusco, Zacatecas, Mexico (21° 44.7′ N, 102° 58.0′ W). The site has a semi-dry climate with summer rainfall, an average annual temperature of 20.3 °C, and approximately 550 mm of annual precipitation. The soil is alkaline, low in organic matter and nitrogen, but high in phosphorus (P) and calcium (Ca) (Padilla-Ramírez et al., 2010, 2014).

A guava germplasm bank has been established at this site, comprising 169 different genotypes collected from 11 Mexican states, as well as from international sources (Padilla-Ramirez et al., 2007, 2014). Each tree is documented with detailed information on its origin, propagation method (seed or cutting), collection location, and collection date. All trees are managed under uniform conditions following the guidelines established by the Secretaría de Trabajo y Previsión Social (NOM-035-STPS-2013, 2013).

Study system

Guava trees (P. guajava) belong to the Myrtaceae family and are distributed from central Mexico to northwestern Argentina (Arévalo-Marín et al., 2021). They are perennial trees that can reach heights of 2–10 m. The species produces solitary, hermaphroditic, actinomorphic flowers with white petals and elliptical leaves. The fruit is globose to ovoid, approximately 8 cm in diameter, fleshy, and variable in color—ranging from white to pink, with some varieties exhibiting red hues. The pulp is soft and juicy, with colorations that include white, beige, yellow, pink, or red (González-Gaona et al., 2002).

Currently, 180 guava genotypes have been registered and are cultivated worldwide, with the greatest diversity found in Mexico, Brazil, India, South Africa, Colombia, Cuba, Thailand, and Malaysia (Padilla-Ramírez et al., 2007, 2014, Arévalo-Marín et al., 2021). In Mexico, the genotypes ‘China’ and ‘Media China’ are the most commercially important, both for domestic consumption and export markets (Martínez-De Lara et al., 2004), and represent the most heavily selected genotypes to date.

Genotype selection

We selected seven guava genotypes from the germplasm collection at the Los Cañones experimental campus in Zacatecas, Mexico. Selection was based on each genotype’s market value and availability within the collection, as most genotypes on site were represented by only a single plant. The selected genotypes were classified into three categories based on their domestication: cultivated genotypes, landraces (“criollas”), and wild genotypes. Cultivated genotypes were originally obtained from commercial orchards that supply national and international markets. Landraces were originally collected from backyard gardens across the country, while wild genotypes originated from mountainous regions with no apparent history of human cultivation (Padilla-Ramírez et al., 2007). Four trees were selected for each genotype, except for ‘Chichimequillas’ and ‘Fuentes,’ for which only three trees were available.

Among the cultivated genotypes, we selected ‘Media China,’ ‘Segregating from India’ (hereafter ‘India’), and ‘54x57.’ ‘Media China’ produces ovoid fruit with yellow skin and cream-colored pulp. Fruit size ranges from 4.0 to 5.0 cm and weighs between 5 and 10 grams. This variety is highly valued in the domestic market (Padilla-Ramírez et al., 2014). The ‘India’ genotype bears round, rough-textured fruit with yellow skin and cream-colored pulp and is notable for its high ascorbic acid and sugar content, which contribute to its nutritional value. The ‘54x57’ genotype produces oval-shaped fruit measuring 5.0 to 6.5 cm in size, with yellow skin and white pulp (Cortés-Penagos et al., 2016; Padilla-Ramírez et al., 2014). Both ‘Media China’ and ‘India’ are currently among the most commercially valuable guava varieties in Mexico.

Two landrace genotypes were included in the study: ‘Robles’ and ‘Mirna-Petra’ (hereafter ‘Mirna’). ‘Robles’ was collected from a backyard near Guadalajara City, Jalisco, and bears oval-shaped fruit with pink pulp. The fruit is occasionally sold by the backyard owners. ‘Mirna’ was collected from backyards in Nayarit State and produces pear-shaped fruit with light pink pulp. Both landraces were sourced from backyards located in or near urban areas.

The wild genotypes selected were ‘Chichimequillas’ and ‘Fuentes.’ ‘Chichimequillas’ was collected from mountainous areas in Guanajuato and produces oval-shaped fruit with yellow skin and white pulp. It has no commercial value. ‘Fuentes’ was collected from the hills of Jalpa, Zacatecas, and is also not used commercially (Padilla-Ramírez et al., 2007). Both wild genotypes were gathered from remote areas without established human settlements.

All plants used in the experiment were managed under uniform agronomic practices, including fertilization, pruning, and pest control, in accordance with the guidelines established by the Secretaría de Trabajo y Previsión Social (NOM-035-STPS-2013, 2013). However, during the year in which measurements were taken, no pesticides were applied to avoid confounding effects. Only fertilization and irrigation were carried out to maintain plant health under standardized conditions.

Leaf morphological traits

We selected four trees per guava genotype, except for the two wild types (‘Chichimequillas’ and ‘Fuentes’), for which only three trees were available in the germplasm bank (N = 26 plants across all genotypes). To evaluate differences in tree size, we recorded tree height and measured vegetative coverage. Vegetative coverage was estimated using the formula π × D 1 × D 2, where D 1 and D 2 are perpendicular canopy diameters encompassing the foliage.

For leaf trait measurements, we collected five mature, undamaged leaf pairs per tree. Leaf pairs were collected due to their morphological arrangement. Leaf area was calculated following the method described by Kobayashi (1988). Leaf thickness was measured using a Fowler MF GID micrometer (0–1 in/25 mm; Fowler High Precision, Newton, MA, USA). Leaf toughness was assessed using two complementary methods: (1) by performing four perforations per leaf with a Chatillon push–pull force gauge (Model 719; Chatillon and Sons, NY, USA), avoiding the major veins, and (2) by calculating leaf mass per area (LMA) using the formula by Witkowski and Lamont (1991):

Fruit morphological traits

Due to differences in fruit availability, we collected 28 fruits from wild plants, 48 from landraces, and 73 from cultivated plants, totaling 149 fruits. All genotypes produced fruit except for the wild ‘Chichimequillas,’ for which we used morphometric data from the previous year provided by INIFAP. For each fruit, we evaluated the following traits: polar diameter (i.e., longitudinal length, in cm), equatorial diameter (i.e., width, in cm), fruit weight (g), pulp thickness (mm), and seed weight (g).

Leaf biochemical traits

To estimate leaf water content, we collected five mature, undamaged leaf pairs per tree (total N = 270). Leaf pairs were collected to obtain sufficient material for chemical analyses. Fresh weight was measured using an analytical balance (Ohaus H-5276, NJ, USA). The leaves were then oven-dried at 40 °C for 48 hours (Binder ED-115, Tuttlingen, Germany) to obtain their dry weight. Water content (WC) was calculated based on the relationship between fresh and dry weights, following the method described by Hellmuth (1970):

To determine the chemical composition of leaves, we collected eight mature leaf pairs per tree (N = 432). The leaves were pressed and sent to the Centro de Innovación Aplicada en Tecnologías Competitivas (CIATEC), where total phenolic compounds were extracted and quantified. Frozen samples were ground in a mortar, and 2 mg of tissue was weighed for polyphenol extraction. Each sample was mixed with 2 mL of 70% methanol and incubated for 12 hours at room temperature on a rotary shaker (250 rpm). After incubation, samples were centrifuged at 10,000 × g for 5 minutes. The resulting supernatants were transferred to new 1.5 mL microcentrifuge tubes and evaporated to dryness using a rotary vacuum evaporator (Büchi Rotavapor R-114 and Büchi Vacobox B-171; Flawil, Switzerland).

Polyphenol quantification was performed using the Folin–Ciocalteu method in a microplate format. Absorbance was measured at 760 nm using a Benchmark Plus microplate reader (Bio-Rad, Hercules, CA, USA). A standard curve was prepared using gallic acid as the reference (Makkar, 2003), and total phenolics are expressed as μg per g of dry tissue.

Fruit biochemical traits

To evaluate total soluble solids (°Brix), a proxy for fruit sugar content, we collected 28, 48, and 73 mature fruits from wild, landrace, and cultivated plants, respectively (N = 149). °Brix was measured using a hand refractometer (Kuhny ATC, Belp, Switzerland).

For the analysis of phenolic compounds in fruits, we collected two mature fruits per tree (total N = 52). Each fruit was analyzed individually and sent to CIATEC, where total polyphenol concentration was quantified using the same extraction and Folin–Ciocalteu assay described above.

Herbivory assessment

To estimate foliar herbivory among genotypes, we first marked five apical leaves on each tree (N = 270). After three weeks, the leaves were collected and analyzed in the laboratory to assess foliar damage using the BioLeaf app (BioLeaf Analysis System, Brazil). The percentage of foliar damage was then calculated. This app is a non-destructive ecological method that uses image processing and computational geometry to calculate total leaf area (reconstructed) and the damaged or missing area (% damage = damaged area ÷ total area), independent of damage type (Brandoli et al., 2016).

Fruit damage was estimated as the proportion of damaged fruits relative to the total number of fruits produced by each tree, with average fruit counts per genotype of 118 for wild, 339 for landrace, and 571 for cultivated plants. Damage caused by the guava weevil (C. dimidiatus) is distinctive, characterized by fruit deformation, a black scar approximately 1 cm² in size, and a characteristic kidney-shaped appearance (González-Gaona et al., 2008). Trees of the ‘Chichimequillas’ genotype had no available fruit and were therefore excluded from the fruit damage analysis.

Statistical analysis

All statistical analyses were performed using RStudio (version 2025.05.0 + 496) (Posit PBC, 2025). Because the experimental unit was the tree, a mean value per tree was calculated prior to analysis. To evaluate the grouping of genotypes based on morpho-biochemical traits, we performed a Principal Component Analysis (PCA) using eight morphological (leaf thickness, foliar area, leaf toughness, fruit weight, mesocarp thickness, polar and equatorial diameter, dry seed weight) and two biochemical (water and sugar content) traits measured in leaves and fruits.

Also, we conducted two sets of analyses: first, we analyzed the data across all genotypes, and then we grouped genotypes into three domestication categories—cultivated, landraces, and wild. Prior to analysis, all variables were tested for normality using the Shapiro–Wilk test and for homogeneity of variances using Levene’s test. To evaluate the effects of guava genotype and domestication category, we applied generalized linear models (GLMs), selecting distributions based on the nature of each response variable. A GLM with a Gamma distribution and log link was used for vegetation coverage per tree, leaf area, LMA, leaf thickness, leaf toughness, water content, and total phenolics in leaves, followed by Dunn’s multiple comparisons test.

For tree height, fruit length, fruit width, total phenolics in fruit, fruit weight, mesocarp thickness, seed weight, and °Brix, we used a GLM with a Gaussian distribution and identity link, followed by Tukey’s multiple comparisons test. Leaf and fruit herbivore damage were analyzed using a GLM with a Poisson distribution and log link, followed by Dunn’s multiple comparisons test. Finally, we used Pearson correlation analyses to assess the relationships among plant morphological and biochemical traits, as well as between herbivory and these traits. In these cases, P -values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method. Only correlations that remained significant after this correction were considered robust.

PCA of morpho-biochemical plant traits

The first two principal components explained 31.7% and 12.2% of the variation, respectively, for a total variation of 43.9% (Figure 2). Some degree of separation was evident for the ‘Robles’ genotype and the genotypes ‘54x57,’ ‘Mirna,’ and ‘Chichimequillas’ (Figure 2). Fruit weight and mesocarp thickness had the highest loading factors (LF) for PC1 (fruit weight LF = 0.45 and mesocarp thickness LF = 0.45), while foliar area and water content contributed most to the variation in PC2 (foliar area LF = -0.57 and water content LF = -0.55).

Leaf morphological traits

Vegetative coverage differed significantly among genotypes (χ² = 4.36, df = 6, P = 0.02); however, there was no significant effect of domestication (χ² = 0.55, df = 2, P = 0.444). The genotypes ‘Robles’ and ‘Media China’ exhibited the highest coverage than the other genotypes (Figure 3).

Neither genotype (F (6,26) = 2.39, P = 0.06) nor domestication (F (2,26) = 0.71, P = 0.501) significantly affected tree height. In contrast, both genotype (χ² = 10.25, df = 6, P < 0.001) and domestication (χ² = 1.08, df = 2, P = 0.003) had significant effects on leaf area. Genotypes such as ‘India,’ ‘Media China,’ and ‘Robles’ produced larger leaves compared to ‘Mirna’ and ‘54x57’ (Figure 4). Cultivated genotypes had larger leaves than wild and landrace genotypes (Figure 4). LMA was not significantly influenced by either genotype (χ² = 0.02, df = 6, P = 0.91) or domestication (χ² = 0.01, df = 2, P = 0.954).

Leaf thickness was significantly affected by genotype (χ² = 11.08, df = 6, P < 0.001) and domestication (χ² = 6.50, df = 2, P < 0.001). Leaves from ‘Chichimequillas’ were thicker than those from ‘India’ and ‘Media China.’ Wild genotypes had the thickest leaves, followed by landraces, and then cultivated genotypes (Figure 5).

Leaf toughness also varied significantly with genotype (χ² = 3.06, df = 6, P < 0.001) and domestication (χ² = 1.77, df = 2, P < 0.001). Among genotypes, ‘Robles’ had the toughest leaves, while ‘54x57’ had the lowest toughness values (Figure 6). Landrace genotypes exhibited the greatest leaf toughness compared to wild and cultivated plants (Figure 6).

Fruit morphological traits

We found significant effects of both genotype and domestication on fruit shape. The genotypes ‘Fuentes,’ ‘Mirna,’ ‘India,’ and ‘Media China’ exhibited greater polar diameters compared to ‘Robles’ and ‘54x57’ (χ² = 8.46, df = 6, P < 0.001; Figure 7A). Across the domestication gradient, two distinct fruit shapes emerged: (1) wild and cultivated genotypes produced pear-shaped fruits, while (2) landraces bore round-shaped fruits (χ² = 4.88, df = 2, P < 0.001; Figure 7A). Genotype (χ² = 2.87, df = 6, P < 0.001) and domestication (χ² = 0.85, df = 2, P = 0.01) also significantly influenced equatorial diameter. Specifically, ‘India’ produced wider fruits than ‘54x57’ and ‘Robles’ (Figure 7B). Cultivated genotypes had wider fruits than landraces (Figure 7B). Considering both diameters, the cultivated genotype ‘India’ produced the largest fruits, while the landrace ‘Robles’ had the smallest.

Fruit weight was significantly affected by genotype (F (6,187) = 32.13, P < 0.001) and domestication (F (2,187) = 19.65, P < 0.001). Among genotypes, ‘India’ produced the heaviest fruits, whereas ‘Robles’ and ‘54x57’ produced the lightest (Figure 8A). Wild and cultivated fruits were heavier than those from landrace genotypes (Figure 8A). Genotype (χ² = 7.2, df = 6, P < 0.001) and domestication (χ² = 19.99, df = 2, P < 0.001) also had significant effects on mesocarp length. The genotype ‘India’ exhibited the longest mesocarp, while ‘Robles’ and ‘Chichimequillas’ had the shortest (Figure 8B). Cultivated and wild genotypes had thicker mesocarps than landraces (Figure 8B).

Seed dry weight was significantly influenced by genotype (F (6,108) = 16.31, P < 0.001) and domestication (F (2,108) = 16.99, P < 0.001). Among genotypes, ‘India,’ ‘Media China,’ and ‘Mirna’ produced heavier seeds than ‘Robles’ and ‘Chichimequillas’ (Figure 8C). Cultivated genotypes had heavier seeds than landraces (Figure 8C).

Leaf biochemical traits

Leaf water content varied significantly among genotypes (χ² = 15.34, df = 6, P < 0.001); however, domestication had no significant effect (χ² = 0.80, df = 2, P = 0.07). ‘Media China’ exhibited the highest leaf water content, while ‘Mirna’ and ‘54x57’ had the lowest values (Figure 9A).

Total phenolic content in young leaves was significantly influenced by both genotype (χ² = 1.06, df = 6, P < 0.001) and domestication (χ² = 1.07, df = 2, P < 0.001). Leaves from ‘Robles’ and ‘Mirna’ contained higher phenolic concentrations than those from ‘54x57.’ Overall, wild and landrace genotypes exhibited higher phenolic levels than cultivated genotypes (Figure 9B). In contrast, total phenolic content in mature leaves was not significantly affected by either genotype or domestication (P > 0.05 for both).

Fruit biochemical traits

Total soluble solids (°Brix) in fruits were significantly influenced by genotype (F (6,187) = 6.17, P < 0.001), but not by domestication (F (2,187) = 2.14, P = 0.12). The genotypes ‘54x57,’ ‘India,’ and ‘Fuentes’ had higher °Brix values compared to ‘Mirna’ and ‘Media China’ (Figure 10).

Domestication significantly affected total phenolic content in mature fruits (χ² = 1.15, df = 2, P = 0.001), with landrace fruits containing 1.3 times more total phenolic compounds than cultivated fruits (Figure 11).

Herbivory assessment

Although overall foliar damage by herbivores was minimal (less than 1%), it was significantly influenced by genotype (χ² = 19.21, df = 6, P = 0.003), but not by domestication (χ² = 0.03, df = 2, P = 0.98). The genotype ‘Mirna’ exhibited more foliar damage than ‘Media China’ (Figure 12A). This damage was primarily caused by leaf beetles, specifically Colaspis sp. (Coleoptera: Chrysomelidae) and Pandeleteius sp. (Coleoptera: Curculionidae) (A.R.-P., personal observation).

Fruit damage caused by C. dimidiatus was not significantly affected by either genotype (χ² = 3.97, df = 6, P = 0.68) or domestication (χ² = 1.38, df = 2, P = 0.499; Figure 12B).

Pearson correlations

After correcting for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR), only the correlation between foliar area and leaf thickness was statistically significant (FDR-adjusted r 2 = -0.274, P = 0.019, Supplementary Table S1). All other correlations were not significant after correction.

No significant correlations were observed between folivory and any leaf morphological or biochemical traits (all P values > 0.05; Supplementary Table S2). Similarly, frugivory showed no significant correlations with any fruit morphological or biochemical traits (all P values > 0.05; Supplementary Table S2).

It is well established that anti-herbivore defenses in plants often come at the expense of other vital functions such as growth and reproduction (Coley et al., 1985; Mooney et al., 2010). Such trade-offs frequently generate genotypic variation—both within and between species—in defensive traits (Fritz, 1992, Agrawal et al., 2002; Gols et al., 2008). This variation is shaped by both natural and artificial selection, as evidenced by divergence between wild and cultivated genotypes under agronomic management. Artificial selection often reduces traits considered less critical for yield. In guava (P. guajava), we found clear genotypic variation in morphological and biochemical traits, although the precise influence of domestication on these traits remains uncertain. Domesticated genotypes generally had larger leaves with reduced thickness, toughness, and phenolic content in young leaves. They also produced larger fruits and seeds with higher sugar and lower phenolic concentrations. While some of these traits also occurred in wild and landrace genotypes, the overall pattern is consistent with a “domestication syndrome,” characterized by enhanced vegetative and reproductive growth at the expense of defenses. Surprisingly, domesticated guavas did not experience higher herbivory, an unexpected finding for a perennial crop native to the Americas. Moreover, we found no evidence of trade-offs between growth, reproduction, and defense, nor between herbivory and any of these traits.

Artificial selection has altered both target and non-target traits in guava. For example, larger leaves in domesticated genotypes often coincided with reduced toughness and phenolic content (e.g., ‘54×57,’ ‘India,’ and ‘Media China’ genotypes) (see Figures 4, 6, 9). Such changes have been linked to improved carbon assimilation, light interception, and nutrient uptake through photosynthesis (Milla et al., 2013; Price et al., 2014). Similar patterns occur in crops such as cabbage, sunflower, chard, wheat, and tomato, where larger leaves are associated with increased vegetative growth (Milla and Matesanz, 2017). Although we did not directly measure photosynthetic efficiency or nutrient uptake, we observed a negative correlation between leaf area and leaf thickness, consistent with a functional trade-off in leaf design. Supporting this, Shiva et al. (2017) reported strong correlations between leaf size and fruit biomass/yield across 33 guava accessions and six Psidium species in India. However, increased nutrient acquisition in domesticated guavas did not correspond to greater tree height or denser foliage. Instead, tree height and canopy coverage were similar across the domestication gradient, likely reflecting standardized orchard practices such as fertilization, pruning, and irrigation (Génard et al., 2008). In Mexico, for instance, the “calmeo” technique involves a 30–40 day post-harvest drought followed by pruning and irrigation to synchronize vegetative and reproductive growth.

Guava domestication patterns resemble those in other perennial crops, where wild traits often persist alongside increased genotypic diversity. Unlike many annuals that may lose wild characteristics entirely, guava has maintained relatively stable morphometric traits across wild, landrace, and cultivated genotypes. Consistent with this idea, the PCA results show that domesticated genotypes exhibit greater variation in the measured traits compared to the other genotypes, which cluster consistently regardless of origin (wild and landraces). Notably, the only genotype that appears to begin differentiating is ‘Robles.’ However, these results should be interpreted with caution, as the total variation explained by the analysis is less than 50%. This suggests guava remains in an early stage of domestication (Martínez-Castillo, 2016), with cultivated forms retaining ancestral features. For instance, fruit traits remain the primary target of artificial selection in guava. In Mexico, farmers typically favor fruit biomass, sugar content above 12°Brix, and high vitamin C concentrations (Padilla-Ramirez et al., 2014), while placing less emphasis on antioxidant content, shape, color, or seedlessness. These preferences likely contribute to the observed variation in fruit morphological and biochemical traits among genotypes. Comparable patterns have been reported in guava studies from both the Americas and Asia (Sanabria et al., 2005; Hernández-Delgado et al., 2007; Mehmood et al., 2014; Singh et al., 2015), indicating parallel selection even across geographically distant regions such as India and Mexico. Unexpectedly, however, fruit traits in wild and cultivated plants were often similar, contrary to our initial hypothesis.

Landraces—intermediate between wild and cultivated forms—can contribute to genetic diversity in domesticated lineages (Villa et al., 2005). Harlan (1975) described landraces as locally adapted populations shaped by traditional agricultural practices, maintaining high genotypic variability. In contrast, cultivated varieties are often bred for reduced variability and greater uniformity. For example, in pitaya cactus (Stenocereus queretaroensis (Weber) Buxbaum), landraces produce larger, sweeter fruits, while the popular ‘mamey’ variety has been bred for peelability (Bernal-González, 2022). In guava, landraces often yield smaller, rounder fruits with bitter flavor and lower sugar content—traits less desirable commercially but important for conserving genetic diversity (Padilla-Ramirez et al., 2007, 2010).

Overall, our morphometric traits in wild guava suggest early artificial selection, with many ancestral features persisting. Historical records indicate guava cultivation in Mexico since pre-Columbian times (Rajan and Hudedamani, 2019), and genetic evidence links Mexican populations to wild plants from the southwestern Amazon—the likely center of origin (Arévalo-Marín et al., 2021). However, the genetic origins of Mexican guava remain uncertain, including whether it was introduced via a single or multiple events during the Spanish conquest. Notably, some genotypes (e.g., ‘Robles’ versus ‘54x57’ and ‘Mirna’) displayed significant divergence regardless of domestication status, whereas the ‘India’ genotype showed the most pronounced domestication-associated traits, including larger leaves and fruits, resembling cultivated types from India (Singh et al., 2015).

Although domesticated plants often exhibit lower phenolic content due to nutrient reallocation (Parker et al., 2010; Whitehead et al., 2017; Whitehead and Poveda, 2019), we found no correlation between phenolic levels (defensive traits) and morphological traits such as leaf or fruit size. This suggests that the reduction in defense traits observed in cultivated guavas may not be solely attributable to resource trade-offs. While foliar herbivory varied among genotypes, domestication history had no detectable effect on herbivory, nor was herbivory correlated with any morphological or biochemical trait. One possible explanation is the coevolutionary history between guava and its herbivores—domestication may not have substantially altered resistance to native pests such as Colapsis spp. and Pandeleteius spp., as seen in other plant–herbivore systems (Hernández-Cumplido et al., 2018). However, it is important to note that domesticated plants exhibited lower concentrations of phenolic compounds in immature leaves. This is an intriguing result because the optimal defense hypothesis predicts that young leaves should be among the most chemically defended tissues for two reasons: (1) they are less protected by physical traits such as toughness or thickness, and (2) they are relatively nutrient-rich compared to mature leaves (McKey, 1974, 1979; McCall and Fordyce, 2010; Bixenmann et al., 2013). An interesting question for future research is whether herbivores associated with guava preferentially feed on young versus mature leaves. Furthermore, the lack of correlation between herbivory and defense or morphological traits in guava suggests that multiple mechanisms may contribute to resistance against native herbivores. Future work should investigate alternative traits—such as flowering phenology, extrafloral nectaries, and volatile emissions—to clarify the mechanisms underlying herbivore resistance.

By growing all genotypes under identical soil, irrigation, and pesticide-free conditions, we minimized environmental variation, while including material from multiple Mexican states and abroad broadened our assessment of guava diversity. Germplasm banks, such as the one used in this study, are essential for such research, enabling controlled trials and safeguarding valuable genetic resources (Padilla-Ramírez et al., 2007, 2010, 2013). Despite their importance, few studies have examined genotypic variability in perennial crops like guava across a domestication gradient under controlled conditions (Parker et al., 2010; Whitehead and Poveda, 2019). Our results provide strong evidence of genotypic variation in guava morphological and biochemical traits, as well as in foliar herbivory. While domestication appears to have reduced certain defensive traits in the leaves and fruits of some genotypes, the extent to which artificial selection has shaped trait expression and herbivore resistance in wild versus cultivated populations remains unclear. Future studies incorporating genomic tools could help disentangle the genetic basis of this variation and provide deeper insight into domestication patterns. Such knowledge is critical for understanding guava’s evolutionary trajectory and for guiding breeding strategies that balance yield, fruit quality, and herbivore resistance. A limitation of our study is that we only used seven genotypes with unclear relationships, which constrains our conclusions regarding domestication. Including additional genotypes would help strengthen our findings.