Introduction

Trichomes are small hairs on the epidermis of stems, leaves, flowers and fruits, present in most plant species (McDowell et al. 2011; Roy et al. 1999). Among other functions, trichomes help reduce transpiration, reflect solar radiation, lower leaf temperature, and attract pollinators (Chang et al. 2016; Glas et al. 2012; Roy et al. 1999; Yang and Ye 2013). Trichomes also contribute to plant defence against pathogens and herbivores through both physical and chemical means, contingent on their morphology and cellular metabolism (Avery et al. 2015; Bergau et al. 2015; Levin 1973; Oney and Bingham 2014; Simmons and Gurr 2005; Tian et al. 2012). For instance, in tomato and its close relatives, up to eight different kinds of trichomes have been described, including glandular (types I, IV, VI, VII) and non-glandular ones (types II, III, V, VIII) (Glas et al. 2012). Within each category, trichomes differ from one another in length (range 0.05–0.30 mm) and number of cells that constitute their base, stalk or gland.

Apart from aiding in water economy, protection from UV radiation, and leaf temperature regulation, non-glandular trichomes may hamper herbivore mobility and foliage consumption (Glas et al. 2012; Kennedy 2003). In Phaseolus, hook-shaped non-glandular trichomes may impale and trap small insects, often causing them to die by starvation (Levin 1973; Simmons and Gurr 2005). In addition, non-glandular trichomes can disrupt the digestive tract of herbivores that ingest them (Kariyat et al. 2017). Meanwhile, glandular trichomes may protect plants against herbivores through the release of a variety of substances upon the rupture of their fragile glands by the feeding action of herbivores. These substances include toxic compounds, resins that immobilize herbivores or gum up their mandibles, and volatiles that can attract natural enemies of herbivores (Degenhardt et al. 2003; Levin 1973). In Solanum alone, over 100 different organic compounds are produced by glandular trichomes, including acyl sugars, terpenes, tannins, alkaloids, mucilage, and resins (Glas et al. 2012; Kennedy 2003; McDowell et al. 2011). The abundant evidence of the negative effects of trichomes on the feeding action of herbivores (Simmons and Gurr 2005 and references therein), clearly places them as important resistance traits of plants, wherein resistance against herbivores is defined as any trait that reduces the preference or performance of herbivores (Karban and Baldwin 1997, emphasis added).

Despite the important role of trichomes in plant defence, few studies have tested hypotheses derived from optimal defence theory on the defensive role of trichomes within the context of ontogenetic or phenological changes in plants. According to optimal defence theory, plant parts or organs with greater fitness value to an individual should be better protected than less valuable parts (McKey et al. 1979). If the fitness value of a leaf is greater when it is a net carbon exporter (older, expanded) than when it is a net carbon importer (younger, unexpanded), then expanded leaves should be better defended than unexpanded leaves (Traw and Feeny 2008, but see; Harper 1989), although this may vary between temperate and tropical species (Coley et al. 1996). Since trichomes are an important component of plant defence systems, the previously stated prediction implies that expanded leaves should have greater trichome density than unexpanded leaves—provided that trichome differentiation is not restricted to a certain period during leaf development.

Similarly, as plants transition from their vegetative growth phase to the production of reproductive structures, resource limitation may result in trade-offs between defence and reproduction (de Jong and Klinkhamer 2005; Koelewijn 2004). Large resource expenditures on reproduction may not allow plants to defend their leaves as efficiently as vegetative plants that are not incurring the costs of reproduction. Accordingly, plants investing in resource-demanding processes such as seed filling, fruit expansion, and embryo growth and maturation, should have leaves with lower trichome density and, consequently, lower resistance to herbivores than those of vegetative plants.

Thus, the goals of this study were (1) to corroborate that trichomes confer resistance against herbivores, and (2) to test the hypotheses that: (a) trichome density and resistance are greater on expanded leaves than on unexpanded leaves, and (b) leaves of reproductive plants at the stage of fruit maturation have lower trichome density and are less resistant to herbivores than those of vegetative plants.

We addressed these hypotheses using tomato as a study system because of its great worldwide importance as a crop, with an average annual global production of 170.4 million tonnes (2012–2016, FAOSTAT 2018), and because of the potential use of trichomes as natural resistance trait that could aid in reducing the dependency of tomato production on synthetic pesticides (Kennedy 2003; Owen et al. 2015; Picanço et al. 2007).

Materials and methods

To assess the relation between trichome density and resistance to a generalist herbivore, in unexpanded and expanded leaves of vegetative and reproductive plants, we grew ten plants from each of six tomato varieties (Solanum lycopersicum L.: Black Cherry [BC], Conquistador, Mistral, Patrón, Roma, and Red Robin [RR]). Two seeds were planted in each 3.8 L pot containing a 4:1 ratio of Pro-mix PGX medium to sand. Pots were placed under compact fluorescent lights with a 16/8 h light/dark cycle at 24 °C and checked daily for germination and adequate soil moisture. Seedlings emerged within a 24 h period 6 days after planting, except for two BC seedlings, which took 1 more day to emerge. Eight days after emergence, plants were thinned down to one individual per pot. All seedlings were grown inside the lab for six more days and later moved to an outdoor roof patio. Pots were placed in six rectangular wire frames, each with 12 spaces, following a stratified randomized block design so that each variety was equally represented in each block. Plants were checked every other day throughout the experiment and watered when the growing medium was dry to the touch. Plants were fertilized weekly with 250 mL 15–15–30 “Plant Prod” fertilizer as per manufacturer directions (2.5 mL fertilizer per 1L water) until mid-September when any remaining fruits were already turning red.

Constitutive resistance was assessed by means of choice bioassays using third- and fourth-instar Trichoplusia ni larvae obtained from Insect Production Services (Natural Resources Canada) and reared on McMorran diet in our lab. To compare resistance between unexpanded and expanded leaves of the different varieties, we used a #5 cork borer to cut out one 0.65 cm2 (0.91 cm diameter) disc from the leaf on the second node from the apex (young, unexpanded) and a second disc from the leaf on the fifth node (expanded) of the same plant, randomly selected from the ten individuals in each variety. This provided us with ten sets of 12 leaf discs (two leaf categories, six varieties). For each of the ten sets of discs, we conducted a feeding trial as follows: five larvae that had fasted for 24 h were placed for 45 min inside a petri plate that contained a set of leaf discs arranged randomly with respect to leaf category and variety, pinned to a foam with moistened filter paper on top. A different set of larvae was used for each feeding trial. At the end of each trial, the remaining area of each disc was measured with the aid of an image analysis system (WinFOLIA™ Pro 2006a, Regent Instruments, Inc., Quebec). The proportion of disc area remaining was used as an estimate of resistance. This operational definition of resistance is consistent with Karban and Baldwin’s (1997) definition in terms of reduced herbivore preference: fully consumed discs (highly preferred) would have the minimum resistance score of zero, indicating low plant resistance, while undamaged discs (least preferred) would have the maximum resistance score of one, indicating high plant resistance.

To compare constitutive resistance between vegetative and fruiting plants, we conducted similar choice bioassays, except that discs were obtained from unexpanded (second node from the apex) leaves collected from plants at two different points in time: first, during the vegetative stage of plants, and later, when plants were producing fruits. On average, the two leaves from a given plant were produced with a difference of 23 days. Using separate sets of plants to have second-node leaves of vegetative and reproductive plants produced concurrently meant each set could have experienced a very different history of exposure to sun, wind, temperatures, and precipitation, among other factors. Instead, to avoid such plethora of potential confounding factors, we used one set of plants, and we froze all leaves as soon as we collected them. In this way, all leaves used for the comparison of resistance between leaves of vegetative and reproductive plants were subjected to freezing, albeit for different period lengths: leaves from vegetative plants remained frozen 34–37 days, and those from reproductive plants, for 11–14 days. When trials were performed, leaves were first thawed for 1 h at room temperature before cutting out discs for the choice bioassays. In parallel with each feeding trial, we placed one leaf disc from each variety on an bioassay arena without larvae for the same time as the experimental discs, and measured their change in area due to desiccation alone. These quantities were later used to correct estimates of disc area remaining and avoid overestimation of the leaf area consumed by larvae.

Because both trichome formation and resistance against herbivores can be induced by damage (Björkman et al. 2008; Karban and Baldwin 1997), we measured trichome density on leaves sampled from a separate set of plants (one per variety) grown concurrently and in the same way as the set of plants used for measurements of constitutive resistance. With the aid of a stereo-microscope (SMZ1500, Nikon Canada Inc., Ontario), we counted all trichomes on the abaxial and adaxial sides of leaf discs cut out using a #1 cork borer (diameter = 0.40 cm; area = 0.13 cm2) from freshly collected leaves of these extra plants in the unexpanded/expanded and vegetative/reproductive leaf and plant stage categories described above. We avoided cutting leaf discs that included the main vein because of the greater trichome densities commonly found along the veins.

As fruits ripened, we collected one healthy-looking fruit from each plant and weighed it (n = 59, one fruit rotted before we could weigh it). Fruits were sliced open, dried to constant weight, weighed and the total number of seeds per fruit were counted. For each plant, we calculated total dry fruit biomass by multiplying the number of fruits produced by the dry mass of the fruit sampled. Similarly, we estimated an individual’s total seed production by multiplying the number of seeds in the sampled fruit by the total number of fruits produced by the plant.

Statistical analysis

We used general linear models (GLM) to analyze the effects of leaf ontogenetic stage (hereafter “leaf stage”) and trichome density on resistance (n = 12). The effect of leaf stage and variety on trichome density was also measured using GLM (n = 12). To analyze the effects of leaf stage and plant phenological stage (vegetative/reproductive) on resistance, first we calculated the differences in resistance between treatments (d1 = Runexpanded − Rexpanded; d2 = Rvegetative − Rreproductive) per individual plant because both leaves for a variety in a petri plate were obtained from the same individual, and then we performed two-tailed Wilcoxon signed rank tests using the null hypothesis that the median difference equaled zero (n = 60). Regression analyses were used to test for a trade-off (negative association) between total seed production per individual and (a) resistance and (b) trichome density. Resistance data were arc-sine transformed (\(y^{\prime}=2 \cdot \sin^{-1} \sqrt y\); (Neter et al. 1990) to attain normality and homoscedasticity of the residuals of the analyses. Graphs show untransformed values for easier interpretation. Minitab 16 Statistical Software® (Minitab Inc., State College, Pennsylvania) was used for all statistical analyses.

Results

We found glandular trichomes types I, VI, and less commonly VII, and non-glandular trichomes types III and V (following Glas et al. 2012). Overall, non-glandular trichomes were over one order of magnitude denser than glandular trichomes (respectively, 3001.0 ± 412.8 [mean ± s. e.], range 230.8–6184.6 and 104.5 ± 16.8, range 15.4–338.5 trichomes cm−2), with greater density of non-glandular trichomes on the abaxial surface (2234.0 ± 306.8) than on the adaxial surface (767.0 ± 120.2), but the opposite for glandular trichomes (abaxial: 29.2 ± 5.0, adaxial 75.3 ± 13.3; all densities in trichomes cm−2). Because herbivores feeding on a leaf encounter and may ingest trichomes from both surfaces and types, here we concerned ourselves with total trichome density.

In vegetative plants, the trichome density of unexpanded leaves was almost six times greater than that of expanded leaves (4037.3 ± 539.1 and 699.8 ± 539.1 trichomes cm−2, respectively; least squares mean ± s. e., hereafter; F1,11 = 19.11, P = 0.007), regardless of variety (F1,11 = 1.90, P = 0.250). We could not assess the effects of the leaf stage-by-variety interaction on trichome density because we had only one observation per leaf stage-variety combination. In addition, young leaves were more resistant to cabbage looper than expanded leaves (0.9977 ± 0.0003 and 0.9892 ± 0.0003, respectively; estimated median difference between unexpanded and expanded leaves: \({\hat {d}_1}\) = 0.036, n = 28, W = 291.5, P = 0.046).

To determine the role of trichomes on the differences in resistance between young and expanded leaves, we tested for the effects of trichome density and leaf stage on resistance. Preliminary inspection of the relation between trichome density and resistance revealed curvature indicative of saturation: a steep increase in resistance with trichome density for densities lower than 2000 trichomes cm−2, but little increase in resistance thereafter (Fig. 1), so we conducted a GLM analysis on ln-transformed variables. We found a strong effect of trichome density (F1,8 = 10.90, P = 0.011), but, interestingly, no significant effects of leaf stage (F1,8 = 0.22, P = 0.653) or the interaction of both factors (F1,8 = 0.51, P = 0.494) on resistance. The lack of a leaf stage effect on resistance in this model, which also includes trichrome density, indicates that these two explanatory variables are aliased to some degree. We used the parameter estimates obtained from the analysis above in the power model \(R=a{\left( {T+1} \right)^b}\), where R is resistance, T is trichome density and a and b are model parameters describing location and curvature of the relationship. The parameter estimates revealed a decelerating curve model: \(\widehat {R}=0.858{\left( {T+1} \right)^{0.019}}\) (notice b < 1; Fig. 1).

Fig. 1
figure 1

Relation between resistance to herbivory by Trichoplusia ni larvae and total trichome density of unexpanded (open circles) and expanded (closed triangles) leaves on vegetative individuals of six varieties of tomato (Black Cherry, Conquistador, Mistral, Patrón, Roma, and Round Robin). Each resistance value plotted represents the mean score of 10 plants per variety × leaf ontogenetic stage combination. Trichome density values were obtained from one individual plant per leaf stage-variety combination, and include the trichomes from both the abaxial and adaxial surfaces of the leaf. The line depicts the best fit according to the power function \(\widehat {R}=0.858{\left( {T+1} \right)^{0.019}}\) (see text for details)

Full size image

We found no significant effects of the phenological stage of the plants (vegetative / reproductive) on resistance (estimated median difference between leaves of vegetative and reproductive plants: \({\hat {d}_2}=0.000\), n = 34, W = 347.0, P = 0.402). (Resistance scores from leaves on vegetative plants subjected to freezing did not differ significantly from those of vegetative plants not subjected to freezing (0.9971 ± 0.0009 and 0.9977 ± 0.0011, respectively, F1,178 = 0.135, P = 0.714). Using the combined trichome density data for vegetative and reproductive plants, we found a significant leaf stage-by-plant phenological stage interaction (F1,15 = 10.050, P = 0.006): expanded leaves of vegetative plants had lower trichome densities than leaves at any other leaf stage-by-phenological stage combination (Fig. 2).

Fig. 2
figure 2

Boxplots of total trichome density on leaves sampled at contrasting ontogenetic stages (unexpanded/expanded) from plants at either the vegetative or reproductive phenological stage. Each boxplot shows the interquartile range (box), the median (thick horizontal line within the box), “whiskers” that extend to the lowest or highest data value within a distance 1.5 times the length of the interquartile range from the lower or upper quartile, and individual values outside that range (open circles)

Full size image

Two varieties, BC and RR, produced the lowest total fruit biomass (F5,51 = 133.7, P < 0.001) and simultaneously many more fruits than the rest (F5,53 = 80.75, P < 0.001, means of 23 ± 2.4 and 31 ± 1.8, compared to 4.7 ± 0.3 to 6.7 ± 0.3). However, we did not find a significant negative association between the number of fruits produced per plant and trichome density (Pearson’s r = − 0.645, d.f. = 4, P = 0.166). In contrast, we found a significant negative association between resistance and the total number of seeds produced per plant (one mean for each variety; Resistance = 0.998–0.000011 seeds; F1,4 = 10.25, P = 0.033; Fig. 3a). Plants of the two least resistant varieties, BC and RR, produced roughly four times as many seeds as plants in any of the other varieties. Interestingly, despite the positive association between trichome density and resistance, we did not detect a negative association between trichome density and seed production (Pearson’s r = − 0.377, d.f. = 4, P = 0.461); BC plants were able to produce both many seeds and many trichomes on their leaves (Fig. 3b). The surprisingly high trichome density found for BC leaves prompted us to check our data against trichome densities obtained for a separate group of plants used in a different experiment, and we found that indeed, BC plants have high trichome densities (7203.3 ± 871.2, mean ± s. e., n = 5). We also tested for a negative association between seed production and trichome density separately for glandular and non-glandular trichomes, and separately in vegetative and reproductive plants, but did not find any statistically significant association (statistical analyses not shown, data on Table S1).

Fig. 3
figure 3

Relation between total seed production per plant and a resistance to Trichoplusia ni, or b trichome density, using mean values for each of six tomato varieties (pooled leaf stage and plant phenological stage for each variety; variety abbreviations as in Fig. 1)

Full size image

Discussion

Our finding that unexpanded leaves of vegetative plants had both greater trichome density and greater resistance to T. ni than expanded leaves indicates that at least part of the resistance to T. ni is effected via trichomes. In fact, adding trichome density as a covariate in the general linear model for the effects of variety and leaf stage on resistance removed the significant effect of leaf stage on resistance, thus indicating that an important amount of the variance explained by leaf stage is in fact due to differences in trichome density. The feeding deterrence function of trichomes is further supported by our finding of an overall positive, diminishing returns curvilinear relation between resistance and trichome density with a steep increase in resistance densities below 2000 trichomes cm−2, but less increase past that density. It must be noted that resistance levels attained by leaves with trichome densities greater than 2000 were very high (Mistral-unexpanded: 0.975 – Patrón-unexpanded: 1.000), which means that larvae of the generalist T. ni almost did not eat from leaf discs with trichome densities greater than 2000 trichomes cm−2. Presently, we do not know whether trichome densities above 2000 trichomes cm−2 deter feeding by specialist insects and, thus are adaptive, or if they are mostly the by-product of a developmental constraint in the production of leaves with high trichome densities (see below). Greater resistance in younger leaves associated to trichome density has also been reported in two species of Brassica (Traw and Feeny 2008).

Overall, leaves showed a high level of resistance (> 0.90) to T. ni, even at very low trichome densities (e.g., 200–300), thus indicating that other characters (e.g., thick cell walls, digestibility reducers, etc.) contribute to reduce the amount of leaf tissue consumed by this herbivore. The positive association between resistance and trichome density found in this study is consistent with other reports on tomato and other species, and corroborates the defensive role of trichomes in plants in general, and specifically in tomato (Handley et al. 2005; Levin 1973; Simmons and Gurr 2005; Tian et al. 2012; Wilkens et al. 1996).

Our results do not support the hypothesized greater trichome density of expanded leaves compared to younger, unexpanded leaves. This hypothesis depended strongly on two assumptions: that expanded leaves have greater fitness value to a plant than unexpanded leaves, and that trichome differentiation is not restricted to early stages of leaf development. Therefore, our finding that unexpanded leaves had greater trichome density and resistance than expanded leaves suggests that expanded leaves do not have greater fitness value than young leaves, and/or that trichomes differentiate only in early stages of leaf development. While it may seem intuitively straightforward that an exporting leaf is of greater value to a plant than a young, unexpanded leaf acting as a net importer of resources at any given time, this instantaneous assessment of fitness values of leaves in different stages of development overlooks the opportunity value of young, unexpanded leaves (Harper 1989)—and therefore, the opportunity costs of losing those structures. As for the second assumption, in some species, trichome differentiation occurs only during early leaf developmental stages (Roy et al. 1999; Werker 2000), and thus, necessarily, as a leaf expands, its trichome density will decrease because the finite number of trichomes differentiated during its unexpanded stage become scattered as regular epidermal cells multiply and increase the area of the leaf. Our results suggest that this is also the case in tomato.

Congruent with both the timing of development of trichomes and the possibility of greater fitness value of young, unexpanded leaves, our results clearly show greater trichome density and resistance against T. ni of young leaves compared to expanded ones in vegetative plants. In this context, it is noteworthy that we did not find a difference in trichome density between young and expanded leaves in reproductive plants. If the allocation of resources to reproductive structures limited trichome production, there should have been an overall decrease in trichome production. Instead, we found similar trichome densities of young and expanded leaves of reproductive plants to those of young leaves of vegetative plants. This unexpected result could be caused by a decrease in the rate of leaf production and expansion brought about by the shunting of resources to reproduction rather than growth. In other words, it is possible that leaves on the fifth node of reproductive plants were not really at the same developmental stage as those on the fifth node of vegetative plants (expanded), but rather at an earlier, less expanded stage.

Trade-offs between defence and reproduction are at the core of evolutionary ecology, particularly resource allocation theory (de Jong and Klinkhamer 2005), but have not been easy to detect (Kornelsen and Avila-Sakar 2015; Stamp 2003; but see; Hoque and Avila-Sakar 2015). In our study, while we did not find evidence of an intra-plant trade-off between defence and reproduction (phenological stage effect on trichome density), at the whole variety level we detected a trade-off between resistance and seed production. BC and RR plants produced many fruits, and therefore, many more seeds than other varieties, but had the lowest resistance levels of all. It would seem that artificial selection for elevated fruit production has come at the expense of resistance in these varieties, as has been suggested would occur in many crop species including tomato (Li et al. 2018; Milla et al. 2015). Such a trade-off also suggests an important cost of resistance, as in Arabidopsis lyrata, in which a trade-off between dry rosette biomass at end of experiment and resistance to the specialist moth P. xylostella has been found (Sletvold et al. 2010).

Curiously, despite the positive association between trichome density and resistance (clearer at lower trichome densities), we did not detect a trade-off between trichome density and seed production because BC plants were able to produce both many seeds and densely pubescent leaves. At this time, we are not sure why RR plants were not able to do so. A more detailed examination of our data revealed that the non-glandular: glandular trichome ratio in expanded leaves of RR vegetative plants (10.1 ± 2.7) was only half that of BC vegetative plants (23.2 ± 8.2), thus indicating that RR vegetative plants invest more heavily on glandular trichomes. Thus, the costs of producing those extra glandular trichomes and synthesizing the diverse secondary compounds made by them could make it impossible for RR plants to produce both densely pubescent leaves and large numbers of flowers, fruits and seeds.

In summary, trichomes are important for resistance against the generalist herbivore, Trichoplusia ni in the six varieties of tomato used in this study. Trichome density and resistance were greater in young, unexpanded leaves of vegetative plants, but we did not see such pattern in reproductive plants. Lastly, we found no effects of the onset of reproduction on trichome density or resistance. However, we detected a cost of resistance expressed as a negative association between resistance and seed production (via fruit production) at the variety level. Thus, it may prove challenging to select for highly resistant varieties that also produce many fruits (and seeds). Instead, we may have to seek varieties with lower fruit production but higher constitutive resistance, or varieties with a high tolerance of herbivory so that they attain an acceptable level of fruit production despite herbivore damage. Future studies should consider whether varieties with low constitutive resistance can attain high induced resistance, and if some of this may be achieved through trichome induction, and also, whether varieties with lower trichome density have greater tolerance to herbivory.