Most studies of trade-offs between defence, growth, and reproduction have examined pairwise correlations between these processes, used ratio-based measures for defence allocation such as allelochemical concentration and trichome density, and estimated resource allocation to growth and reproduction in terms of biomass. However, for statistical and biological reasons, it may be preferable to analyse these processes holistically, to use absolute amounts of resistant traits and leaf mass/area, and to measure growth and reproduction in terms of nodes with or without flowers. 

We aimed to identify how leaf stinging hairs as resistant structures and leaf area as a functional trait affected growth and reproduction in the Japanese stinging nettle Urtica thunbergiana. 

We conducted a greenhouse experiment with nettles derived from a population that has been historically exposed to heavy browsing by Sika deer (Cervus nippon) in Nara Park, Japan. We analysed causal relationships between stinging hair number, leaf area, growth rate, growth performance, and reproductive output using structural equation modelling (SEM). In this analysis we adopted newly developed indices for a plant’s stinging hair number and leaf area, and measures of growth and reproductive traits in terms of nodes instead of biomass. 

There was a significant covariation between stinging hair number and leaf area. Stinging hair number had direct negative effects on growth rate and performance, while leaf area had positive direct effects on growth rate, growth performance and reproductive output. The growth rate had a significant direct positive effect on reproductive output, resulting in a significant indirect negative effect of stinging hair number on reproductive output.

This indicates that there is not only a trade-off between resistance and growth, but also an indirect trade-off between resistance and reproduction through reduced growth rate and suggests that U. thunbergiana sacrifices growth rather than reproduction to increase resistance. Our study provides future work on trade-offs between defence, growth, and reproduction with a new methodological framework that can assess indirect as well as direct trade-offs, together with the effects of leaf area as a functional trait on these processes.

In our data set, the sampling unit of leaf was nested within plant. We therefore constructed a two-level hierarchical linear model (HLM) to examine the relationship between leaf area and stinging hair number at the within-plant and between-plant levels:

Within-plant level

Yij = β_0j + β_1j Xij + r_ij

Between-plant level

β0j = γ₀₀ + γ₀₁ G_j + u_0j

β1j = γ₁₀ + γ₁₁ G_j + u_1j

where

Y_ij = stinging hair number for leaf i of plant j

X_^ij = leaf area for leaf i of plant j

G_j = mean leaf area of plant j

β_0j = plant-specific intercept

β_1j = plant-specific slope

γ₀₀ = overall mean intercept (fixed effect)

γ₁₀ = overall mean slope (fixed linear effect of leaf area)

γ₀₁ = regression coefficient associated with mean leaf area of plant relative to plant-specific intercept (fixed linear effect of mean leaf area of plant)

γ₁₁ = regression coefficient associated with mean leaf area of plant relative to plant-specific slope (fixed effect of the interaction between leaf area and mean leaf area of plants)

u_0j = random effect of plant j on plant-specific intercept

u_1j = random effect of plant j on plant-specific slope

r_ij = residual

In this model, X_ij was centred at G_j, while G_j was centred at the grand mean value. The parameters γ₁₀ and γ₁₁ represent the effects of leaf area at the within- and between-plant levels, respectively. We then tested the statistical significance of the fixed and random effects. In this procedure, the parameters were estimated by maximum likelihood and robust standard errors were calculated. The parameters γ₀₀, γ₀₁, γ₁₀, and γ₁₁ and the variance components of u_0j and u_1j were tested using t-test and likelihood ratio test, respectively.

Structural equation modelling analysis (SEM)

We conducted SEM to examine the effects of leaf area and stinging hair number on growth rate, growth performance, and reproductive output as follows. Firstly, we built the full model under the hypotheses that (1) leaf area and stinging hair number covary; (2) both leaf area and stinging hair number affect the growth and reproductive components; (3) growth rate affects both growth performance and reproductive output; and (4) growth performance also affects reproductive output. Secondly, we constructed multiple models by leaving or removing insignificant (P > 0.05) paths from the initial model. Next, we computed the χ2 measure of goodness of fit (except for the full model due to lack of degrees of freedom) and the corrected Akaike information criterion for small sample sizes (AICC) for the full and each improved model. We then selected the model with an insignificant χ2 and the lowest AICC as the best model that fit the data better than any of the other models. Finally, we evaluated the relationships between the variables on the basis of the path coefficients of the best model.