*P<0.05, **P<0.01, ***P<0.001
^{a}Heritability for lifehistory strategy was estimated by dividing the genotypic variance from generalized linear mixed models by the total phenotypic variance according to the binomial equation: total # plants x frequency of annual plants x frequency of rosette plants. The value in parentheses shows the heritability from REML where the 0/1 lifehistory variation was treated as a continuous response.
Table 2. Plant traits that explain
variation in herbivory by Popillia japonica. Stepwise regression identified one lifehistory (LH) trait (annual/biennial) and four phenolic traits that explained 73% of variation in the proportion of leaves damaged by an outbreaking exotic herbivore. No significant quadratic effects were detected. A multivariate distillation of these traits using principal components analysis explained 45% of the variation. Note: The first principal component (PC1) contrasted concentrations of kaempferol glucuronide, flavonoid glycoside (
11), total ellagitannins and total phenolics (positive loadings) with quercetin glucuronide and a flavonoid glycoside (
9) (negative loadings). PC 2 provided a
contrast between bolting date, % water, and oenothein B, and two ellagitannins (
4,7) (all with positive loadings) with biomass, C:N, oenothein A, an ellagitannin (
1), oenothein A, other ellagitannins and total phenolics (negative loadings). The regression coefficient (parameter),
Pvalue,
and partial r^{2} values are shown for each trait and PC; the sum of partial
r^{2} is >1 because some variables were collinear. Bolded numbers in parentheses denote peaks from HPLC (see Fig. 1).
Trait

Parameter

Pvalue

r^{2 }(partial)

LH (annual/biennial)

0.012 ± 0.006

0.06

0.10

Ellagitannin (1)

0.007 ± 0.003

0.06

0.10

other ellagitannins

0.001 ± 0.0003

0.001

0.29

quercetin glucuronide (8)

0.011 ± 0.004

0.01

0.19

other flavonoids

0.173 ± 0.039

<0.001

0.38





PC1

0.013 ± 0.006

0.03

0.12

PC2

0.04 ± 0.008

<0.001

0.42

Table 3. Genotypic selection gradients describing directional (
β), nonlinear (
γ) selection and selection differentials (S) on
O. biennis traits. The full multiple regression model explained 97% of the total variance in relative fitness among 39
O. biennis genotypes (genotypes were treated as replicates). Because of extensive covariance among variables (see Gmatrix, Supplementary Table 1) we also used principal components analysis to reduce the data to a smaller number of variables.
After stepwise regression, five PC’s explained 79% of the variation in lifetime fitness among genotypes. Note: PC1 contrasted total phenolics, total ellagitannins, kaempferol glucuronide and a flavonoid glycoside (
11) (all with positive loadings) with quercetin glucuronide and a flavonoid glycoside (
9) (all negative loadings). PC2 contrasted the positive
loadings of bolting date, % water, oenothein B, and an ellagitannin (
4) with the negative loadings of herbivory,
plant biomass, C:N, oenothein A, other ellagitannins and total phenolics. PC3 contrasted the positive loadings of lifehistory strategy,
trichome density, oenothein B, other flavonoids with the negatively loaded caffeoyl tartaric acid. PC5 contrasted LH strategy (positive loading) with bolting date (negative loading). Bolded numbers in parentheses correspond to HPLC peaks (see Fig. 1).
Trait

β
 P 
γ
 P 
S


biomass

0.58 ± 0.11

<0.001

0.37 ± 0.12

0.003

0.92***


LH strategy

0.47 ± 0.08

<0.001

0.24 ± 0.14

0.09

0.84***


oenothein A (5 and 6)

0.18 ± 0.07

0.01

0.05 ± 0.06

0.41

0.62***


quercetin glucuronide (8)

0.13 ± 0.05

0.02

0.15 ± 0.07

0.03

0.20


Principal components 





PC1

0.44 ± 0.21

0.04





PC2

2.05 ± 0.25

<0.001





PC3

1.23 ± 0.28

<0.001





PC5

0.79 ± 0.37

0.04





Fig. 1. Characteristic HPLC traces of phenolics found in
O. biennis leaves. (
A) Chromatogram from a plant sample with a high concentration of oenothein B and a low concentration of oenothein A, and (
B) a sample with a low concentration of oenothein B and a high concentration of oenothein A. The numbered compounds were quantified individually: (
1) ellagitannin, (
2)
caffeoyl tartaric acid, (
3) oenothein B, (
4) ellagitannin, (
5 and
6)
isomers of oenothein A, (
7) ellagitannin, (
8) quercetin glucuronide, (
9)
flavonoid glycoside, (
10) kaempferol glucuronide, and (
11) flavonoid glycoside.
Fig. 2. The distribution of genetic correlations and associations among functional traits in
O. biennis. A) The frequency distribution of pairwise Pearson genetic correlations among all traits (
N = 276 pairwise correlations), where the mean
rvalue is shown by the dashed vertical line. B) Associations among traits according to hierarchical cluster analysis using Ward’s linkage. We identified four clusters of traits separated by sumsofsquares distance >2; average pairwise correlations within and between clusters are shown in parentheses.
Fig. 3. Plant traits that explained variation in herbivory by A)
Popillia japonica. the major herbivore observed on plants (photo credit: A. Agrawal). Multiple regression using the genotype BLUPs revealed that genetic variation in concentrations of: B) quercetin glucuronide, C) “other flavonoids”, and D) “other ellagitannins” explained the most variation in herbivory (see Table 1). Genetic variation in lifehistory strategy and an ellagitannin (peak 1; see Fig. 1) also explained variation in herbivory, but these traits explained less variation. All figures show residual herbivory versus
residual trait variation, where variation in other traits has been partialed out.
Fig. 4. Directional and quadratic selection on
O. biennis. Multiple regression using the genotype BLUPs revealed four plant traits that explained 97% of the variation in relative fitness among plant genotypes. A) Strong directional selection acted to increase plant biomass in plant genotypes with low biomass, but stabilizing selection acted at high biomass values. Directional selection also acted to increase: B) the initiation of flowering
from the rosette stage, and C) the concentrations of an ellagitannin (peak 1). D) Directional selection acted to decrease concentration of quercetin glucuronide in genotypes with low concentrations of this compound, and we detected significant disruptive selection with a fitness minimum at moderately high concentrations. All figures show residual fitness versus residual trait variation, where variation in other plant traits was partialed out.
Fig. 1.
Fig. 2.
Fig. 3.
F
ig. 4.
Supplemental Table 1. Matrix of variances, covariances and genetic correlations among functional plant traits.
(see attached Excel file)
Supplemental Table 2. Selection differentials on each of twentyfour plant traits. Selection differentials were measured according to Price (1970), by calculating the covariance between the BLUPs of relative fitness and the normally standardized genotypic breeding values (BLUPs). Bolded numbers in parentheses refer to HPLC peaks (Fig. 1)
Trait

S

Pvalue

Herbivory



proportion of leaves damaged

0.65

<0.001

Foliar traits

0.79

<0.001

C:N

0.21

0.21

SLA

0.54

<0.001

% water content

0.46

0.003

trichomes



Phenolics (mg/g dry tissue)



Total phenolics

0.66

<0.001

caffeoyl tartaric acid (2)

0.00

0.99

Ellagitannins (ET)



et (1)

0.28

0.08

oenothein B (3)

0.26

0.11

et (4)

0.21

0.2

oenothein A (5 and 6)

0.62

<0.001

et (7)

0.04

0.82

other et's

0.60

<0.001

total et

0.52

0.001

Flavonoids (Flav)



quercetin glucuronide

0.20

0.21

flav (9)

0.37

0.02

kaempferol glucuronide (10)

0.58

<0.001

flav (11)

0.53

0.001

Other flav's

0.41

0.009

Total flav

0.65

<0.001

Lifehistory and phenology



strategy (rosette0/flowered1)

0.84

<0.001

bolting date (days since Feb. 1)

0.59

<0.001

flowering date (days since Feb. 1)

0.01

0.94

biomass (g)

0.92

<0.001
