Heritabilities and response of iridoids to selection
In the present study, we obtained an upper estimate of 0.84 for the heritability of leaf iridoid glycoside content in the leaves of P. lanceolata, based on a full-sib analysis of the base population. As this estimate includes an unknown dominance component, it should be regarded as an upper estimate of the narrow-sense heritability of this trait. Estimates of the realized heritability (±SE) based on four generations of artificial selection were 0.23 ± 0.07 and 0.23 ± 0.04 for upward and downward selection, respectively. This indicates that under greenhouse conditions, the genetic component of variation in constitutive levels of iridoid glycosides is 23%. The response declined during consecutive generations; estimates based on the first generation only were two and three times higher than the average response over four generations for downward and upward selection, respectively. This decline can only be partly explained by the expected decrease in genetic variance known as the Bulmer effect ( Falconer & Mackay, 1996; p. 203). Insight into possible causes of this decline would require further study. The estimate of 0.23 is within the range of heritability estimates for secondary plant metabolites found in other studies. For instance, narrow-sense heritabilities for furanocoumarins in P. sativa ranged from 0.07 to 0.91 ( Zangerl et al., 1989 ), estimated heritabilities for shoot and root pyrrolizidine alkaloids in S. jacobaea were 0.33 and 0.43 ( Vrieling et al., 1993 ; Van Dam & Vrieling, 1994), for 2′-cinnamoyl salicortin and salicortin in S. sericea 0.59 and 0.20 ( Orians et al., 1996 ), for phenolic resin content in D. aurantiacus 0.32 ( Han & Lincoln, 1994), and for glycosinolates and their myrosinase enzyme in B. rapa 0.17 and 0.35, respectively ( Siemens & Mitchell-Olds, 1998). The presence of the significant heritable component of variation in leaf iridoid glycoside levels in P. lanceolata indicates that selection imposed by generalist insect herbivores could potentially result in an adaptive response of leaf iridoid levels in the field.
The response of iridoid glycoside levels to selection resulted in a four-fold difference in iridoid glycoside content in the leaf pairs that were under direct selection between the upward and downward selected lines. Both components of the total iridoid glycoside content, aucubin and catalpol, responded to selection, but the relative increase of catalpol during upward selection exceeded that of aucubin. The ratio of catalpol to the total iridoid glycosides therefore increased in the H line, whereas it remained relatively constant in the L and in the M lines. Some studies have shown that catalpol is more deterrent to generalist insect herbivores than aucubin ( Puttick & Bowers, 1988). The shift towards a higher proportion of catalpol may therefore result in an even stronger deterrence of plants selected for high levels of iridoids than expected on the basis of their total iridoid content. Selection on leaf iridoid glycosides was accompanied by responses of iridoid glycoside levels in other plant parts. Differences in iridoid glycoside content between the H and L lines were not only observed in the leaf pairs that were subject to direct selection but also in the other leaves, although the magnitude of the response in those leaves was smaller than in the leaves under direct selection. The 2.5 fold higher shoot iridoid level in plants from the H line compared with plants from the L lines clearly indicates that the response to selection was not merely caused by allocating a higher or lower fraction of a total amount of shoot iridoids to the leaf pair subject to selection. Selection on high leaf iridoid glycoside levels also led to an increase in stalk iridoid glycoside content. Our results therefore indicate that iridoid glycoside levels in leaf and stalk tissues are genetically correlated. This suggests that if leaf feeding insects impose selection on host plants for higher leaf iridoid levels, the response of the host plants to these herbivores may be accompanied by increased resistance towards nonadapted generalist insects feeding on reproductive tissues.
Responses of morphological traits to selection
Although we selected on the levels of a secondary plant compound, a suite of morphological traits showed a correlated response, suggesting that there are genetic correlations between physiological and morphological traits. Such genetic correlations between secondary plant compounds and other phenotypic traits have also been found in other studies. In P. sativa, furanocoumarin levels in the seeds showed significant positive genetic correlations with seed length, width, and the number and area of vittae ( Zangerl et al., 1989 ).
A striking result was that selection for high leaf iridoid glycoside content produced plants with few but large leaves, few side rosettes with a corresponding low number of leaves, and few stalks, whereas L line plants showed a consistent pattern in the opposite direction. The observed differences in growth form between plants from the H and L lines are remarkably similar to differences between two previously described ‘hayfield’ and ‘pasture’ ecotypes, respectively, in P. lanceolata ( Van Groenendael, 1986; Wolff, 1987; Van Tienderen & Van der Toorn, 1991a, b; Van Hinsberg & Van Tienderen, 1997). Hayfield ecotypes (subvar. latifolia Wimm. et Grab.; Pilger, 1937) are characterized by rosettes with few but long, erect leaves and inflorescences. They rarely produce side rosettes and their inflorescences have elongated spikes and long scapes. Pasture ecotypes (subvar. sphaerostachya Mert. et Koch f. minor;Pilger, 1937) form small prostrate rosettes with many small leaves and ascending inflorescences with small roundish spikes. Hayfield ecotypes are found in high, dense vegetation such as hayfields whereas pasture ecotypes occur in open, often grazed vegetation. Reciprocal transplant studies have shown local adaptation of these ecotypes to their natural habitats ( Van Tienderen & Van der Toorn, 1991a), which may be partly related to the ability of hayfield plants to intercept more light in their competitive native habitat with a high and dense vegetation, and to the higher incidence of flowering and maturation of spikes of pasture plants in their native habitat ( Van der Toorn & Van Tienderen, 1992). Studies by Wolff (1987), Wolff & Van Delden (1987, 1989) and Van Hinsberg (1996) have shown that there are strong genetic associations between traits involved in these two complexes. For instance, upward selection on leaf length under simulated sunlight (high R/FR-ratio) ( Van Hinsberg, 1996) and downward selection on leaf angle ( Wolff & Van Delden, 1989) in P. lanceolata produced plants showing a suite of traits consistent with the hayfield ecotype. In the present study, the plants used in the crosses to produce seeds for the base population were collected from two populations that were neither explicitly hayfield nor pasture types. But even if they would have been, the emergence of hayfield and pasture types when selecting on iridoid glycoside content is not self evident. Apparently, leaf iridoid glycoside content is genetically correlated with the complex of traits underlying these two growth forms.
The association between growth form, including leaf production, and iridoid level raises the question whether differences in iridoid levels between plants from the H and the L lines simply reflect an inherent allocation pattern of iridoids that is linked to development, or that there are development-independent differences in iridoid levels between these lines. In agreement with the general pattern observed for within-plant distribution of secondary plant compounds ( Zangerl & Bazzaz, 1992), previous studies have shown that levels of iridoid glycosides in P. lanceolata decrease with leaf age, either because of reallocation or because of breakdown of these compounds in older leaves (e.g. Bowers & Stamp, 1992). This means that in plants with a high rate of leaf production (e.g. that have produced 18 leaves within 5 weeks), the iridoid level in an ontogenetically defined leaf stage (such as the third and fourth leaf pair) may be low simply because these leaves are relatively ‘old’ compared with the third and fourth leaf pair produced on plants with a lower rate of leaf production (e.g. that have produced only 14 leaves in that period of time). Indeed, in the base population, at a fixed plant age, a significant part (39.9%) of the variation in iridoid level among full-sib families could be explained by the average number of leaves produced in the main rosette ( Fig. 4A, P < 0.001). Families that had produced on average 18 leaves in the main rosette, had only half the concentration of iridoids in their third and fourth leaf pair compared with families that had produced on average only 14 leaves in the main rosette ( Fig. 4A). Nevertheless, the response to selection could only be partly ascribed to such developmentally related variation in iridoids. During subsequent generations, the among-family correlation between number of leaves and iridoid level disappeared within lines ( Fig. 4B,C), and there was a large difference in iridoid level between plant families with a similar number of leaves between the two lines. Apparently, continued selection also resulted in an increase in the average level of iridoids in plants with a high rate of leaf production in the H line, and in a decrease in the average iridoid level in plants with a low rate of leaf production in the L line. Analysis of plants from the third generation (the last generation for which replicates at the family level were available) using leaf number as a covariate ( Table 5), confirms that a large part of the differences in iridoid level among families from the H and L lines are developmentally independent (F[1,96]=282.5, P < 0.001). Apparently, the selection response has been based on both development-related and on development-independent variation in iridoid levels.
Figure . 4. Level of iridoid glycosides (% dw) (A–C) and the ratio of catalpol to the total iridoid level (D–F) in the third and fourth leaf pair, as a function of the number of leaves in the main rosette. Data points are family mean values for full-sib families of the base population (A and D), and half-sib families of the first (B and E) and third generation (C and F). Open and closed triangles: family mean values for the H and L lines, respectively. Dotted lines: regressions using all family mean values; solid lines: within-line regressions. Squares: population mean values.
Download figure to PowerPoint
Table 5. The ANOVA of the effects of selection line and leaf number in the main rosette (covariate) on the level of iridoid glycosides (ln-transformed) and the ratio of catalpol to the total iridoids in the third and fourth leaf pair of P. lanceolata. Analyses based on family mean values for female half-sib families from lines selected for H and L levels of leaf iridoid glycosides that constituted the third generation of selection.
By contrast, one of the correlated responses, the higher ratio of catalpol to the total iridoids in the H line, mainly reflected developmental differences between the two lines ( Fig. 4F, Table 5). Old leaves generally contain a lower ratio of catalpol to total iridoids than new leaves (e.g. Bowers & Stamp, 1992). Accordingly, plants with a high rate of leaf production tended to have a lower catalpol-to-total ratio in the relatively ‘old’ third and fourth leaf pair ( Fig. 4D, P < 0.10). In the third generation, differences in the catalpol-to-total iridoid ratio between families appeared to be mainly due to differences in the number of leaves in the main rosette ( Fig. 4F, P < 0.001). Differences between families from the H and L lines that were independent of development were small ( Fig. 4F, Table 5, F[1,96]=4.93, P=0.029).
The association between iridoid level and growth form has a number of interesting consequences. First, we expect that this association has consequences for plant attack by generalist and specialist insect herbivores in natural populations. From our results, we would predict that hayfield plants show higher overall levels of iridoid glycosides in their leaves than pasture plants. This may result in a higher overall deterrence of hayfield plants to nonadapted generalist insects than pasture plants, but higher palatability to adapted specialists that use iridoid glycosides as feeding or oviposition stimulants ( Bowers, 1991). These predictions will be tested in the near future. Second, a genetic correlation between low leaf iridoid glycoside levels and traits associated with the ‘pasture’ ecotype could in principle impose constraints on the simultaneous evolution of resistance to generalist insects (by increased levels of secondary metabolites) and to larger grazing herbivores (by a high production rate of small procumbent leaves and inflorescences) in open, grazed habitats. In that case we would predict that in open, grazed habitats, patches that suffer higher pressure from generalist insect herbivores show increased levels of iridoid glycosides, associated with a shift towards more ‘hayfield’ like traits.
In conclusion, we have found significant heritable variation for leaf iridoid glycoside content in P. lanceolata. The levels of iridoid glycosides in the leaves are genetically correlated with the levels of these compounds in the reproductive tissues and with plant growth form. These results suggest, first, that selection for high levels of these compounds in the leaves by generalist, leaf feeding insects may result in increased resistance to generalist insects that feed on reproductive tissues. Secondly, in open, grazed habitats, selection for increased levels of these compounds in the leaves may be hampered by their association with a set of morphological traits that has previously been shown to be unfavourable in this type of habitat.