Introduction

Elevational gradients can result in large variations in environmental conditions over relatively short distances (Körner 2007). Therefore, such gradients can be used in natural experiments to simulate climate change scenarios (Khairiyah et al. 2013; Buckley et al. 2019). Studies of elevational gradients and insect-plant interactions have primarily addressed woody vegetation and perennial herbaceous plants in natural environments (reviewed by Moreira et al. 2018). Inferences drawn from such studies, often conducted in natural but disparate environments, may be confounded by differences in herbivore populations, abiotic conditions, or plant phenotypic plasticity (Ohsawa and Ide 2008; Rokaya et al. 2016). Common garden experiments under controlled climatic conditions may provide a better understanding of plant defense plasticity and its effect on insect-plant interactions (Rokaya et al. 2016; Shah et al. 2020).

At higher elevations, a low-level of herbivore abundance and plant damage are predicted due to harsher environmental conditions (Salgado et al. 2016; Wu et al. 2021). Such reduction in herbivory pressure is expected to reduce the level of constitutive plant secondary metabolites (PSMs) leading to a lower level of plant resistance (Dostálek et al. 2016; Kergunteuil et al. 2018; Pellissier et al. 2014; Robert et al. 2019). However, widespread applicability of this prediction is problematic; upon closer inspection, there are cases that do not fit the pattern. In some cases, defense investments were enhanced with elevation (Buckley et al. 2019; Galmán et al. 2019), while others have found no or non-linear associations (Buckley et al. 2019; Pellissier et al. 2014). These inconsistencies, from studies conducted mostly on trees and herbs, are indicative of a complex interaction between insects and plants in relation to elevation. Unfortunately, there is limited knowledge of variation in plant defense responses along elevation gradients in agricultural systems highly sensitive to both biotic and abiotic stressors (Smilanich et al. 2016; Smith and Beaulieu 2009).

Elevation also affects the level of induced PSMs in response to herbivore damage. In pine species (Family: Pinaceae), the level of induced resins was reduced with elevation, whereas constitutive defenses were elevated (Moreira et al. 2014). A similar reduction in inducibility of glucosinolates (GLS) with elevation in Cardamine species was reported by Pellissier et al. (2016). In contrast, induction of GLS was stronger in mid- and high-elevation populations of Arabis alpina (Family: Brassicaceae) (Buckley et al. 2019). It is important to note that these studies analyzed different kinds of PSMs, which are affected by herbivory pressure and also by an adaptive response to different environmental stresses (e.g., ultraviolet, temperature, CO2) (Moreira et al. 2018). PSMs, such as total phenolics content (TPC), flavonoids, proline, tannins, and ascorbic acid were higher in leaves from higher elevation plants in several endemic plants from different families, including the Lamiaceae, Rosaceae, and Caryophyllaceae (Hashim et al. 2020). Induction of jasmonate-induced proteins (e.g., protease inhibitors) in response to adverse abiotic factors at higher elevations was also reported in the same study. TPC in leaves was enhanced at higher elevations in Plantago lanceolate (Family: Plantaginaceae) and Medicago minima (Family: Fabacae), presumably as a protective strategy against high ultraviolet radiation (Kabtni et al. 2020). Similarly, increased leaf trichome densities in plants from higher elevation were reported for Nepeta heliotropifolia (Family: Lamiaceae; Talebi et al. 2019), Artemisia tridentata sp. Wyomingensis (Family: Asteraceae; Dyslin 2012), accessions of Solanum berthaultii (Family: Solanaceae; Horgan et al. 2009), and several other flowering plants (Wu et al. 2021). High-density trichomes protect plants from UV-B radiation and heat loss at higher elevations (Karabourniotis et al. 1998; Kofidis and Bosabalidis 2008). Knowledge of plant defense responses in plants modulated by a combination of herbivore pressure and abiotic factors should help improve predictions of plant growth under climate warming.

In the present study, we tested the hypothesis that the level of constitutive PSM and defensive proteins are higher in tomato accessions originating from a lower elevation and therefore strengthen plant resistance against herbivory (Buckley et al. 2019; Dostálek et al. 2016). Conversely, the strong inducibility of these defensive compounds in tomato accessions from higher elevations has been hypothesized (Rasmann et al. 2014). Plant responses to changes in biotic (insect pests) and abiotic effects (environmental variations) were assessed by measuring defensive proteins [polyphenol oxidase (PPO), trypsin protease inhibitors (TPI)], TPC, leaf trichome density, and herbivore bioassays. PPO and TPI play important roles in enhancing plant defenses in tomato by interfering with the digestive physiology of insect herbivores (Felton et al. 1989). The activity of TPI, PPO, and phenolics are affected by biotic and abiotic stresses in tomatoes (Lin et al. 2021; Paudel et al. 2020). Two wild relatives of tomato, Solanum lycopersicum var. cerasiforme (SLC) and S pimpinellifolium (SP), both native to the Andes in South America (Blanca et al. 2015), were used in this study. Experiments were conducted under common-garden conditions to control the abiotic habitat (Garrido et al. 2012).

Materials and Methods

Plant Materials

Tomato seeds were collected from the Andes Mountains in South America and obtained from the C.M. Rick Tomato Genetics Resource Center, UC Davis, Davis, CA, USA (https://tgrc.ucdavis.edu). Three accessions each of SLC and SP originated from a broad elevation range (100 m to 3000 m; Table 1) and were used for the experiments. One accession each from SP (LA 0400) and SLC (LA 2213) were sourced from a single plant; the rest of the accessions were from multiple individuals/populations. As both SP and SLC are of South American origin, we assumed that the plants used in the experiment are the product of generations of exposure to their respective elevational gradient (R. Chetelat, TGRC, personal communication, Sept 10, 2019). Plants were grown in a greenhouse environment (27 °C ± 2 °C, 60–70% RH, 16:8 L:D) at Penn State University campus in central Pennsylvania, USA. Seedlings were grown in Metromix 400 potting mix (Premier Horticulture, Quakertown, PA, USA) and watered every 1–2 d until the end of the experiment.

Table 1 Tomato accessions and original locations of wild tomato (Solanum pimpinellifolium) and cherry tomato (S. lycopersicum var. cerasiforme) genotypes over a broad elevation range in South America (source: Tomato Genetics Resource Center at University of California, Davis, USA)
Full size table

Insect Herbivore

Eggs of Helicoverpa zea were purchased from Benzon Research (Carlisle, PA, USA). Neonates were reared individually inside a plastic cup (30 ml) on a wheat germ and casein-based diet (Bioserv, Frenchtown, USA) until used for the experiment. Fifth instars were used to damage leaflets, whereas 1st instars were used for bioassay experiments (details below).

Defensive Proteins

Two groups of plants for the two treatments (damaged and undamaged) were randomly assembled from plants growing in the greenhouse. Constitutive and induced levels of two jasmonic acid (JA)-related defensive proteins, polyphenol oxidase (PPO) and trypsin protease inhibitor (TPI), were measured in undamaged and caterpillar-damaged leaves. At the four-leaf stage, fully expanded terminal leaflets were used as the focal leaves for defensive protein bioassays (Tan et al. 2018). Leaves were damaged, allowing a 5th instar H. zea to feed inside a clip cage (3.15 cm2) attached to a leaflet. Average time to consume the entire section of leaf tissue inside the clip cage was between 2 and 3 h. An empty clip cage from a separate plant was used as a ‘control’. Following the completion of feeding, the insect and the clip cage were removed from the plant. Leaf tissues (50 mg) from both damaged and control leaflets were collected 48 h later using liquid nitrogen and stored at −80 °C until analysis. Tissues from damaged leaves were collected around the edges bordering the herbivore-damaged area. Similar portions of leaf tissues were also collected from the control. PPO and TPI activities were measured using a spectrophotometric method described by Acevedo et al. (2017). PPO activity was determined based on reaction with caffeic acid and measured as the change in absorbance at 450 nm over 5 min. The values were then normalized by the amount of protein in the sample and expressed as mean optical density (mOD)/min/mg protein. Bradford assay, using Coomassie brilliant blue, was used to quantify protein content in leaf samples (Bradford 1976). TPI activity was estimated using an inhibition rate for the reaction between trypsin and p-toluene-sulfonyl-L-arginine methyl ester and calculated as % inhibition/mg protein. The protocol to measure TPI activity is described in detail by Zhao et al. (2019). The inducibility of defensive proteins was estimated by dividing the level of induced activity by the constitutive levels and expressed as fold change.

Total Phenolics Content

TPC was analyzed in both damaged and undamaged leaves using a colorimetric assay as described by Ainsworth and Gillespie 2007. Briefly, leaf tissues (20 mg) were harvested and stored as described above for defensive proteins. Leaf tissues were homogenized, and 1.5 ml ice-cold methanol (95%) was added and allowed to extract for 48 h in the dark. Samples were centrifuged and F-C reagent and Na2CO3 added to 100 ml of the supernatant. Absorbance was recorded at 765 nm using a spectrophotometer and the amount of total phenolics was calculated (μg/mg tissue) using gallic acid equivalents. Inducibility was estimated by dividing the level of induced activity by the constitutive levels and expressed as fold change.

Density of Leaf Trichomes

The youngest terminal leaflets from randomly selected three-week old tomato seedlings were used to compare the density of trichomes on the adaxial leaf surface (Paudel et al. 2020). Two leaf discs (0.6 cm diam.) were punched out from each side of the mid-vein of a leaflet. The density (number/cm2) of all glandular and non-glandular trichomes was determined by counting all trichomes under a light microscope.

Herbivore Bioassays

Herbivore feeding bioassays with excised leaves (damaged or undamaged) in a laboratory were used as a proxy for measuring plant resistance. Leaves were damaged using 5th instar H. zea as described above for defensive proteins. At 48 h post-damage, randomly selected 1st instars (n = 11–14/treatment) from a stock colony were weighed (‘day 0’) and placed into plastic cups (30 ml) with damaged and undamaged leaves under ambient temperature. Individual larvae were weighed 48 h later (‘day 2’) and the relative growth rate (RGR) was calculated as:

$$\mathrm{RGR}\ \left(\mathrm{weight}\ \mathrm{gain}/\mathrm{g}/\mathrm{day}\right)=\left({\mathrm{W}}_2-{\mathrm{W}}_0\right)/\left[\left({\mathrm{d}}_2-{\mathrm{d}}_0\right)\ast {\mathrm{W}}_0\right].$$

Wo and W2 are larval weight at days d0 and d2 (Waldbauer 1968).

Statistical Analysis

Activities of plant defensive proteins (PPO and TPI), TPC, and RGR of insect larvae were analyzed using a two-way ANOVA with the main effects being ‘tomato accessions’ and ‘insect treatment, damaged or undamaged, plus all interaction terms (Minitab Inc. 2018). Trichome density data were analyzed using one-way ANOVA with ‘tomato accessions’ as the main effect. PPO activity, TPI activity, and trichome density data were LOG10 transformed to meet the normality assumption; data were back transformed for presentation.

The relationships between RGR and PPO, TPI, or TPC were examined using linear regression (Minitab Inc. 2018). The average values for RGR, PPO, TPI and TPC were calculated for each tomato accession originating from different elevations and two treatments (damaged vs. undamaged), resulting in 12 data points (6 accessions × 2 treatments).

Results

Defensive Proteins

Activity of TPI in leaves from both tomato genotypes (SLC and SP) did not vary with elevation origin (SLC: F2, 35 = 0.30, p = 0.75, SP: F2, 35 = 0.07, p = 0.93), but was affected by insect damage (SLC: F1, 35 = 153.80, p < 0 .001, SP: F1, 35 = 33.90, p < 0.001) (Fig. 1a, b). Interactive effects of elevation origin and insect damage were significant for SLC accessions (F2,35 = 5.80, p < 0.001), but not for SP accessions (F2,35 = 0.4, p = 0.68).

Fig. 1
figure 1

Activity of trypsin protease inhibitor (TPI) in undamaged and damaged leaves from four-leaf stage tomato plants: a Solanum lycopersicum var. cerasiforme (SLC) accessions collected from three different elevations, LA 1286 (150 m), LA 2213 (800 m) and LA 4708 (2800 m), and b Solanum pimpinellifolium (SP) accessions collected from three different elevations, LA 0375 (100 m), LA 0400 (800 m), and LA 1547 (3000 m). Fifth instar Helicoverpa zea were used to damage leaves. Bars are mean ± SEM. Means with different letters are different (as determined by Tukey HSD, P < 0.05) among tomato accessions, whereas * denotes a difference between the damaged and undamaged plants within individual tomato accessions

Full size image

Among SLC accessions, LA 2213 (800 m) had the highest TPI activity in undamaged leaves, whereas LA 1286 (150 m) had the highest post-herbivore damage. Inducibility of TPI was highest for LA 1286 (150 m) with a 3.3-fold increase, followed by LA 2213 (800 m; 2.0-fold increase) and LA 4708 (2800 m; 1.9-fold increase). Among SP accessions, LA 0375 (100 m) had the highest TPI activity in undamaged leaves, whereas LA 0400 (800 m) had the highest post-herbivore damage. Inducibility of TPI was highest for LA 1547 (3000 m) with a 2.6-fold increase, followed by LA 0400 (800 m; 2.1-fold increase) and LA 0375 (100 m; 1.8-fold increase).

Activity of PPO varied between SLC and SP accessions originating from different elevations. There was an effect of elevation origin within SLC accessions (SLC: F2,35 = 5.20, p = 0.01) but not for SP accessions (SLC: F2,35 = 1.70, p = 0.20) (Fig. 2a, b). Insect damage affected PPO activity in leaves (SLC: F1,35 = 133.40, p < 0.001, SP: F1,35 = 80.00, p < 0.001). Interactive effects of elevation origin of genotypes and insect damage were not significant for both SLC and SP accessions (SLC: F2,35 = 1.90, p = 0.20, SP: F2,35 = 1.40, p = 0.30).

Fig. 2
figure 2

Activity of polyphenol oxidase (PPO) in undamaged and damaged leaves from four-leaf stage tomato plants: a Solanum lycopersicum var. cerasiforme (SLC) accessions collected from three different elevations, LA 1286 (150 m), LA 2213 (800 m) and LA 4708 (2800 m), and b Solanum pimpinellifolium (SP) accessions collected from three different elevations, LA 0375 (100 m), LA 0400 (800 m), and LA 1547 (3000 m). Fifth instar Helicoverpa zea were used to damage leaves. Bars are mean ± SEM. Means with different letters are different (as determined by Tukey HSD, P < 0.05) among tomato accessions, whereas * denotes a difference between damaged and undamaged plants within individual tomato accessions

Full size image

Among SLC accessions, LA 4708 (2800 m) had the highest PPO activity in undamaged leaves, whereas LA 1286 (150 m) had the highest post-herbivore damage. Inducibility of PPO was highest for LA 2213 (800 m) with a 11.5-fold increase, followed by LA 1286 (150 m; 9.3-fold increase) and LA 4708 (2800 m; 5.2-fold increase). Among SP accessions, LA 0375 (100 m) had the highest PPO activity in both damaged and undamaged leaves. Inducibility of PPO was highest for LA 1547 (3000 m) with a 20.5-fold increase, followed by LA 0400 (800 m; 10.8-fold increase) and LA 0375 (100 m; 8.3-fold increase),

Total Phenolic Content

TPC varied with elevational origin in both SLC and SP accessions (SLC: F2,29 = 9.12, p = 0.001, SP: F2,30 = 0.51, p = 0.001) (Fig. 3a, b). Insect damage, however, did not affect TPC activity in leaves (SLC: F1,29 = 3.19, p = 0.085, SP: F1,30 = 2.96, p = 0.096). Interaction effects of elevation origin of genotypes and insect damage were not significant for both SLC and SP accessions (SLC: F2,29 = 0.29, p = 0.75, SP: F2,30 = 2.02, p = 0.15).

Fig. 3
figure 3

Total phenolics content (μg/mg tissue) in undamaged and damaged leaves from four-leaf stage plants: a Solanum lycopersicum var. cerasiforme (SLC) accessions collected from three different elevations, LA 1286 (150 m), LA 2213 (800 m) and LA 4708 (2800 m), and b Solanum pimpinellifolium (SP) accessions collected from three different elevations, LA 0375 (100 m), LA 0400 (800 m), and LA 1547 (3000 m). Bars are mean ± SEM and means with different letters are different (as determined by Tukey HSD, P < 0.05) among tomato accessions

Full size image

Among SLC accessions, LA 4708 (2800 m) had the highest activity of TPC in both damaged and undamaged leaves. Inducibility was highest for LA 2213 (800 m) with a 1.12-fold increase, followed by LA 1286 (150 m; 1.06-fold increase) and LA 4708 (2800 m; 1.03-fold increase). Among SP accessions, LA 1547 (3000 m) had the highest level of TPC in both damaged and undamaged leaves. Inducibility of TPC was highest for LA 0375 (100 m) with a 1.22-fold increase, followed by LA 0400 (800 m; 1.04-fold increase) and LA 0375 (100 m; 1.0-fold increase),

Density of Leaf Trichomes

On both SLC and SP accessions, elevation origin affected leaf trichome density on undamaged leaves (SLC: F2,60 = 3.90, p < 0.01, SP: F2,60 = 12.70, p < 0.001) (Fig. 4a, b). Trichome density in SP accessions was highest for LA 1286 (150 m) followed by LA 2213 (800 m) and LA 4708 (2800 m). Among SP accessions, the highest trichome density was observed in LA 2213 (800 m) followed by LA 1286 (150 m) and LA 4708 (2800 m).

Fig. 4
figure 4

Density of glandular and non-glandular trichomes (number of trichomes/cm2) on undamaged leaf surfaces (adaxial): a Solanum lycopersicum var. cerasiforme (SLC) accessions collected from three different elevations, LA 1286 (150 m), LA 2213 (800 m) and LA 4708 (2800 m), and b Solanum pimpinellifolium (SP) accessions collected from three different elevations, LA 0375 (100 m), LA 0400 (800 m), and LA 1547 (3000 m). Bars are mean ± SEM and means with different letters are different as determined by Tukey HSD

Full size image

Herbivore Performance

On both SLC and SP accessions, RGR varied with elevation origin on undamaged leaves (SLC: F2,78 = 8.10, p < 0.01, SP: F2,78 = 17.70, p < 0.001) (Fig. 5a, b). Previous insect damage affected RGR in SLC accessions (F1,78 = 5.30, p = 0.02) but not for SP accessions (F1,78 = 2.20, p = 0.13). Among SLC accessions, RGR of H. zea was lowest on undamaged leaves from LA 4708 (2800 m), followed by LA 1286 (150 m) and LA 2213 (800 m). When fed on previously damaged leaves, RGR was lowest on LA 4708 followed by LA 2213 and LA 1286. The percent reduction in RGR was comparatively higher on damaged leaves from LA 2213. On average, RGR was reduced by 1.34-fold, 1.14-fold, and 1.04-fold on damaged leaves compared to undamaged leaves from LA 2213, LA 4708, and LA 1286, respectively. Interactive effects of elevation origin and insect damage were not significant for both genotypes (elevation × insect damage; SLC: F2,78 = 1.50, p = 0.23, SP: F2,78 = 0.06, p = 0.94).

Fig. 5
figure 5

Relative growth rate (RGR) (mass gained/g/day) of first instar Helicoverpa zea fed on detached leaves (damaged or undamaged): a Solanum lycopersicum var. cerasiforme (SLC) accessions collected from three different elevations, LA 1286 (150 m), LA 2213 (800 m) and LA 4708 (2800 m), and b Solanum pimpinellifolium (SP) accessions collected from three different elevations, LA 0375 (100 m), LA 0400 (800 m), and LA 1547 (3000 m). Fifth instar H. zea were used to damage leaves and the bioassay was conducted 48 h post-damage. Bars are mean ± SEM. Means with different letters are different (as determined by Tukey HSD, P < 0.05) among tomato accessions, whereas * denotes difference between damaged and undamaged plants within individual tomato accessions

Full size image

Among SP accessions, RGR of H. zea was lowest on both undamaged and damaged leaves from LA 1547 (3000 m), followed by LA 0400 (800 m) and LA 0375 (100 m). The reduction in growth was also comparatively higher on damaged leaves from LA 1547. On average, RGR was reduced by 1.11-fold, 1.08-fold, and 1.05-fold on damaged leaves compared to undamaged leaves from LA 1547, LA 0375, and LA 0400, respectively. In both SLC and SP accessions, interactive effects of elevation origin and insect damage were not significant.

Statistical differences were not observed between the two wild tomato species (SLC and SP) based on linear regression analysis of RGR with TPI, PPO, and TPC. Therefore, the relationship between RGR and TPI, PPO, and TPC are presented using a single regression line regardless of the two wild tomato species (Fig. 6a–c). RGR of H. zea was negatively correlated with all three defensive chemicals (TPI, PPO, TPC) and was significant for TPC (F1,10 = 1.50, p = 0.002).

Fig. 6
figure 6

Relationship between relative growth rate (RGR) (mass gained/g/day) and defensive chemicals based on linear regression analysis on Solanum lycopersicum var. cerasiforme (SLC) and Solanum pimpinellifolium (SP) accessions: a Trypsin protease inhibitor (TPI) b Polyphenol oxidase (PPO), and c Total phenolics content (TPC). S. lycopersicum var. cerasiforme (SLC) accessions collected from three different elevations, LA 1286 (150 m), LA 2213 (800 m) and LA 4708 (2800 m), and b S. pimpinellifolium (SP). The two wild tomato species were not different; therefore, relationships are presented using a single regression line between RGR and defensive chemicals. Arithmetic value of regression line obtained from the analysis along with the p value are presented. Each dot represents average values of RGR and defensive chemicals for individual tomato accessions; SLC accessions are presented in black round circles, whereas red squares denotes SP accessions

Full size image

Discussion

The elevational origin of tomato plants affected both host plant defenses and herbivory. Tomato accessions from mid and high elevations were most resistant against H. zea, whereas those originating from low elevation were most susceptible. There was a substantial variation in level of constitutive and induced PSMs between two tomato genotypes and among accessions within each genotype. This suggests a complex plant response to varying levels of herbivore pressure and environmental factors along elevational gradients.

Growth of H. zea was reduced on both damaged and undamaged leaves from higher elevation accessions for both tomato genotypes (SLC and SP). The level of TPC in leaves was higher on plants from higher elevations, which negatively affected herbivore growth rate, as documented in other systems (e.g., Oak, Quercus robur, Abdala-Roberts et al. 2016; Himalayan Yellow Sage, Salvia nubicola, Dostálek et al. 2016). The results, however, contrast with the general prediction of relatively less defensive investment at higher elevations due to decreasing herbivory rates (Moreira et al. 2018; Rasmann et al. 2014). This indicates that herbivore pressure alone is inadequate for explaining variations in plant defense investment strategies along elevational gradients. Therefore, the effect of changing abiotic conditions on plant defenses is likely an important consideration (Robert et al. 2019). For example, phenolic compounds were higher at high elevations (as found in the present study), presumably to protect plants from an increased level of UV radiation (Close and McArthur 2002). Increases in the levels of other PSMs, such as cyanogenic glycosides (CNglcs) and glucosinolates, at higher elevation have also been reported in Bird’s-foot trefoil, Lotus corniculatus (Salgado et al. 2016), and Cardamine species (Pellissier et al. 2016), respectively.

Elevational trends in constitutive and induced levels of defensive proteins were not apparent. However, tradeoffs between two defense strategies were evident for both tomato genotypes in accordance with previous studies (Kempel et al. 2011; Rasmann et al. 2014). Mid and high elevation SLC accessions exhibited the highest levels of constitutive TPI, PPO, and TPC. In contrast, the inducibility of these defenses was highest for low and mid-elevation accessions. Among SP accessions, higher levels of constitutive TPI and PPO were recorded in low elevation plants, whereas TPC was highest in plants originating from high elevations. Conversely, strong inducibility of TPI and PPO were recorded in high elevation plants, whereas TPC was highest in plants originating from the lowest elevation. Similar asymmetric patterns were also observed with leaf trichomes.

Concluding remarks and future directions

Variations in host plant defense responses to elevation, as observed in this study, may have important ecological consequences in the context of climate change. Such variations indicate plasticity in host plant responses adapted to different elevations or locations. As insect species expand their range further from the equator, the strength of genetic-based host plant defense mechanisms may play an important role in either restricting or facilitating range expansion (Marta et al. 2019; Shah et al. 2020). Further, crop losses are predicted to rise in the face of global climate change (Deutsch et al. 2018); therefore, variation in plant-defensive traits from plant genotypes/accessions adapted to different climates may be utilized to produce resilient agricultural crops (Paudel et al. 2019). Future studies should consider involving reciprocal transplants along a continuous elevation gradient at one or more sites to support the results and conclusions of this study. Similarly, multi-trophic interactions with a complete set of insect pests associated with a crop are necessary to further our understanding of how biotic and abiotic ecological factors, in combination with evolutionary processes, affect ecosystem dynamics.