Herbivores can have contrasted impacts on litter quality and litter decomposition, through an alteration of leaf chemistry and leaf senescence. Depending on the context, herbivores can induce defensive secondary compounds and thus slow down litter decomposition or accelerate decomposition by short-cutting nutrient resorption.
Almost nothing is known for grasses, which contain smaller amounts of secondary compounds than forbs and trees. Because grasses span a gradient from exploitative species having a low C : N ratio and induced defences, to conservative species having a high C : N ratio and constitutive defences, we hypothesize that the litter dynamics of functionally contrasted grasses may be differentially altered by herbivores.
In a mesocosm experiment, we assessed the litter decomposition rate of two subalpine grasses, the more exploitative Dactylis glomerata and the conservative Festuca paniculata, in the presence of two grasshopper species, Chorthippus scalaris and Euthystira brachyptera. We hypothesized that decomposition patterns depending on grass species and herbivory could be explained by the C : N ratio and the total phenolic content of fresh, senescent and decomposed leaves.
Herbivory by grasshoppers induced the accumulation of phenolics in the fresh leaves of D. glomerata, but most of these compounds were lost during senescence. The decomposition rate of D. glomerata senescent leaves did not depend on herbivory, phenolics and N content or C : N ratio. In contrast, herbivory did not induce any phenolic accumulation in the grazed leaves of F. paniculata, but during senescence, phenolics disappeared in greater proportions in grazed leaves than in ungrazed leaves, probably due to the physical alteration of grazed leaves. Herbivory slowed down the decomposition rate of F. paniculata, which was correlated to the phenolic concentration of senescent leaves, but not to the C : N ratio or N content.
Herbivory by grasshoppers differentially altered the litter decomposition rate of the two functionally contrasted grasses, having no effect on D. glomerata and slowing down F. paniculata. Thus, the combination of chemical and physical modifications of leaves by grazing and their interaction with grass traits may have either accelerating or decelerating effects on litter decomposition, with potentially complex outcomes at the ecosystem level.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Below- and above-ground linkages mediated by the effects of herbivores on plants and soil processes are recognized as a major driver of terrestrial ecosystems (Frank & McNaughton 1993; Bardgett & Wardle 2010). Herbivores can affect soil processes either directly, through their frass, their cadavers and the alteration of soil parameters like temperature and moisture, or indirectly, through their effects on plant community composition and on the quantity and quality of plant litter and root exudates (Hunter 2001; Bardgett & Wardle 2003).
The alteration of litter quality deserves specific attention because it is one of the most complex below- and above-ground linkages. Indeed, herbivory can induce different types of secondary compounds in plants such as polyphenols, flavonoids and phenolic acids (Oksanen et al. 1987; Cipollini et al. 2008; Ibanez, Gallet & Després 2012), with potentially different consequences for litter decomposition (Findlay et al. 1996; Schimel, Cates & Ruess 1998; Schweitzer et al. 2005). For example, tannins could slow down decomposition, while phenolic acids have been reported to exert an accelerating effect (Schimel, Cates & Ruess 1998; Hättenschwiler et al. 2011). Also, herbivory by insects can alter litter quality and decomposition through a limitation of nutrient resorption in leaves, which accelerates decomposition (Chapman et al. 2003; Sariyildiz et al. 2008). Moreover, Chapman, Schweitzer & Whitham (2006) proposed that herbivory induces recalcitrant compounds in deciduous trees, while it limits nutrient resorption in conifers, so that the herbivory–litter quality linkage depends on the plant taxa considered. Hence, the complexity of herbivory–litter quality linkage may be responsible for the observation that responses of litter decomposition to herbivory are idiosyncratic (Bardgett & Wardle 2003; Schweitzer et al. 2005; Chapman, Schweitzer & Whitham 2006; Frost et al. 2012).
To date, almost all the studies that track the influence of herbivores on litter quality and litter decomposition have focused on woody plants, but little specific work has been conducted on grasses. Studies that investigated the effects of grazing-induced grass species replacement suggest that herbivory tends to reduce litter quality and slows down its decomposition (Semmartin et al. 2004; Garibaldi, Semmartin & Chaneton 2007). However, the effects of herbivory on litter quality and decomposition of grazed vs. ungrazed grasses are almost unexplored. In a study with large grazing herbivores, Semmartin, Garibaldi & Chaneton (2008) found that grazing either enhanced or retarded litter decomposition of two functionally contrasted grass species (Lolium multiflorum and Paspalum dilatatum) and hypothesized that the contrasting N economy and photosynthetic metabolism of the plants involved might explain this idiosyncrasy. Indeed, grasses are functionally heterogeneous, as are trees. There is a contrast between exploitative species, which are fast growing with a low C : N ratio and conservative species, with opposite traits (Grime 1977). Additionally, secondary compounds might be at play in grasses, in a similar fashion to deciduous trees. Although grasses contain no tannins (Barbehenn & Peter Constabel 2011), they contain relatively small amounts of phenolics and relatives such as phenolic acids and flavonoids (Sánchez-Moreiras, Weiss & Reigosa-Roger 2003). Moreover, grasses cover a broad functional range from conservative species that tend to rely on constitutive defences like toughness and silica to exploitative species able to accumulate secondary compounds (Cebrian & Duarte 1994; Rosenthal & Kotanen 1994).
We therefore hypothesize that (i) herbivory of grasses by insects affects litter decomposition through the alteration of both the C : N ratio and the amount of secondary compounds in leaves and that (ii) functionally contrasted grasses are impacted by herbivory in different ways. To test these hypotheses, we manipulated herbivory by grasshoppers in experimental communities (mesocosms) of two grass species with contrasting traits and compared the litter decomposition of consumed and unconsumed grasses. We then explored the chemistry of the fresh, senescent and decomposed leaves, in order to explain observed decomposition patterns.
Materials and methods
Study Area, Plant and Insect Species
The experiment was conducted at the Station Alpine Joseph Fourier, at the Col du Lautaret (2100 m) in the central French Alps. All plants and grasshoppers were collected in the neighbouring subalpine grasslands, on the south-facing aspect of the commune of Villard d'Arène, ranging from 1700 to 2100 m (see Quétier et al. 2007 for a detailed description of the site and its vegetation). As model plants, we used two dominant perennial tussock species, that is, the conservative Festuca paniculata (L.) Schinz & Thell., which has tough leaves and high C/N, and the exploitative Dactylis glomerata (L.), which has tender leaves and low C/N (Gross, Suding & Lavorel 2007). As model herbivores, we used two Gomphocerinae grasshoppers, the habitat generalist Chorthippus (Stauroderus) scalaris Fischer–Waldheim, present in most plant communities in the study area, and the habitat specialist Euthystira brachyptera Ocskay, restricted to tall-grass meadows and shaded environments.
In early May 2010, single tillers of both species were separated from large tussocks and planted in individual godets filled with potting soil. One month later, in early June 2010, they were planted in large pots (Ø 45 cm, h 50 cm) filled with a standard mixture of 2/3 of sand, 1/6 of vermiculite and 1/6 of potting soil and fertilized with 3 g of low leaching rate fertilizer (Fertiltop, 16 − 8 − 10 + 4 MgO + oligoelements) in order to reduce nutrient limitation. In each plot, we planted 24 individual plants, according to the following design:
‘D’ treatment: 21 D. glomerata and three F. paniculata individuals (i.e. 7 : 1)
‘F’ treatment: three D. glomerata and 21 F. paniculata individuals (i.e. 1 : 7)
‘50’ treatment: 12 D. glomerata and 12 F. paniculata individuals (i.e. 1 : 1)
The plants mixtures were designed for the purpose of a companion study focusing on the effects of herbivory on plant–plant competition (Ibanez et al. in press). One month later, the pots were covered by an insect nylon mesh at 50 cm above the ground and four adult grasshoppers were introduced as follows in each of the plant composition treatments:
‘Cs’ treatment: two males and two females of C. scalaris
‘Eb’ treatment: two males and two females of E. brachyptera
‘CsEb’ treatment: one male and one female of each species
‘0’ treatment (control): no grasshoppers.
Grasshopper density corresponded thus to 25 individuals m−2, which is about 5–10 times the observed natural densities in the surrounding communities (S. Ibanez, pers. obs.). In total, 48 mesocosms were set, with four replicates for each combination of plant composition and herbivory treatments (three plant treatments, four grasshopper treatments, 48 = 3 × 4 × 4). Mesocosms were checked weekly, and the dead grasshopper bodies removed and replaced by living individuals of the same species and sex, in order to keep the density and proportion of the species, and thus herbivory pressure constant. The grasshoppers were present in the mesocosms from the 29 June 2010 until the 27 September 2010 and from the 27 June 2011 until the 10 September 2011 when the experiment was stopped. The grasshoppers were removed earlier in 2011 to allow the harvest of fresh leaves before the beginning of senescence. The same mesocosms were studied during these two consecutive years. All the individual plants had several herbivory marks at the end of each season, so that the herbivory pressure was substantial enough to affect leaf chemistry. Grasshoppers were not selective regarding plant species, they consumed D. glomerata and F. paniculata according to their relative availabilities in the mesocosms, but due to the greater tolerance of D. glomerata, herbivory increased the competitive ability of D. glomerata with respect to F. paniculata (Ibanez et al. in press).
On 2 November 2010 (about 4 months after the introduction of the grasshoppers), senescent leaves on both plant species were harvested from the plants in each mesocosm and dried for 48 h at 40 °C before the decomposition experiment. One hundred milligram of senescent leaves of each species (D. glomerata and F. paniculata) was used for the D treatment and F treatment, respectively. For the 50 treatment, we used 50 mg of both plant species. The bulked senescent leaves were then put in 5 cm × 4 cm litter bags with distinct mesh size, large mesh size (1500 μm) and small mesh size (68 μm, 68PES4/135; DIATEX, St-Genis-Laval, France), for the purpose of another experiment focusing on the relative contribution of microbial and animal decomposers. In order to disentangle the potential effects of herbivory on the quality of senescent leaves and on the mesocosms’ soil environmental conditions, the litter bags were placed on the ground at two distinct places, namely in the mesocosm in which the senescent leaves were harvested (experimental conditions) and in a neighbouring natural grassland dominated by F. paniculata (field conditions). Therefore, for each mesocosm treatment, four litter bags were designed, while both mesh size (small vs. large) and location (mesocosm vs. field) factors were combined for a total number of 192 litter bags (48 × 2 × 2). The litter bag experiment started mid-November 2010 and ended mid-May 2011. At this subalpine site, there is substantial winter time decomposition, especially under a thick snowpack (Saccone et al. 2013). The decomposed leaves were extracted from the litter bags and dried for 24 h at 40 °C. The percentage of mass loss during the November–May interval was calculated as: % mass loss = 100× (mass before decomposition − mass after decomposition)/mass before decomposition.
For each mesocosm, an additional sample (10 mg) randomly sampled from the senescent leaves harvested for the decomposition experiment was dried for 48 h at 40 °C only, because higher temperatures could alter phenolic compounds and kill the microbes present at the leaf surface that might be involved in decomposition (Graça, Bärlocher & Gessner 2005). Senescent leaves were then ground to powder and analysed for carbon and nitrogen concentration using a CHN analyser (CHS NA1500; Carbo Erba Instrument, Milan, Italy). At the end of the decomposition experiment, 10 mg of the decomposed leaves of each litter bag was analysed in the same way. Ten milligram of fresh leaves harvested in August 2011 was also analysed for carbon and nitrogen concentration. Fresh leaves were collected at the very end of the experiment to avoid artificial herbivory during the experiment.
Leaf phenolics were extracted twice in a 70 : 30 ethanol–water solution under reflux at 100 °C. After filtration, the two extracts were bulked and the final volume measured. The total phenolic compound concentration was determined with Folin–Ciocalteu reagent after evaporation under vacuum. The total phenolic concentration was calculated by comparison with a calibration curve made with gallic acid and expressed in milligram of gallic acid equivalent per gram of dry mass.
Separation and UV characterization of the different monomers of phenolic compounds were performed using high-performance liquid chromatography (HPLC) coupled with diode array detector. Injection volume was 20 μL on RP C18 column at 1·5 mL min−1 flow. Phenolic compounds were separated depending on their hydrophilic character and molecular weight. Two different gradient methods were used depending on the phenolic class. For low-molecular-weight phenolic monomers (including phenolic acids and aldehydes characterized by one aromatic ring), the percentage of acetonitrile in acetic acid (at 0·5% in distilled water) ranged from 0% to 20% during 45 min. For flavonoids, the percentage of methanol in acetic acid (at 0·5% in distilled water) ranged from 10% to 60% during 30 min. The HPLC technique provides for each sample a chromatogram of several peaks that are recorded at the maximum wavelength of absorbance (see Appendix S1 in Supporting Information for examples of chromatograms). Each peak corresponds to a single compound or a group of related compounds. Preliminary analysis revealed that the flavonoid and phenolic acid contents in senescent and decomposed leaf extracts were too low to be measured, and we thus restricted our analysis to fresh leaves.
We used the free r software (R Development Core Team 2011) for all analyses. The significance threshold was fixed to 0·05. The C : N ratios of the fresh, senescent and decomposed leaves were square-root transformed prior to analysis. Total phenolics were log-transformed to achieve normality. Preliminary analyses showed that the different combinations of grasshopper species (Cs, Eb and CsEb treatments) gave similar results (data not shown). Thus, to gain analysis power, we defined the herbivore factor as the presence (either Cs, Eb or CsEb treatments) or absence (O treatment) of grasshoppers.
The effect of the presence of grasshoppers on leaf parameters was tested with type III anova tests with the presence/absence of grasshoppers and leaf state (fresh, senescent and decomposed) as factors. We used the ‘anova’ function from the ‘car’ r package (Fox & Weisberg 2011), which is suitable for unbalanced designs. When the anovas were significant, for each leaf response variable, the pairwise statistical differences between factor levels (i.e. the presence/absence of grasshoppers and the leaf state, fresh, senescent, decomposed) were tested with Tukey's tests using the glht function of the multcomp r package (Hothorn, Bretz & Westfall 2008). The heterogeneity of variances was taken into account with the ‘sandwich’ r package (Zeileis 2006). We calculated the percentage of variation of total phenolics during senescence as follows, with Px standing for the concentration of phenolics in leaf state ‘x’: % variation = 100× (Psenescent − Pfresh)/Pfresh. We calculated the percentage of variation during decomposition similarly, using Pdecomposed and Psenescent. No statistical tests were made on these percentages because they would have been redundant with the tests described above.
We evaluated the influence of plant species identity and of the presence/absence of herbivores on the percentage of mass loss in the litter bags. We also checked whether the location and mesh size of the litter bags interacted with the plant and herbivore treatments. We included the four factors (plant species, herbivores, location, mesh size) as well as the interactions in 25 mixed models (Bates, Maechler & Bolker 2011) with mesocosms as a random factor. Mixed models were required here because there were several litter bags for each mesocosm, whereas in the leaf chemistry analysis, there is a single sample per mesocosm. The 25 models ranged from an intercept model (one parameter) to the model including all factors and all interactions (24 parameters, Table 1). For each model, we calculated the Bayesian Information Criterion (BIC), the Akaike Information Criterion (AIC) and the corrected AIC (AICc). The model having the lowest BIC and AICc values was considered as the best model, in the sense that it included the factors and their interactions that best explained litter mass loss. Nonadditive effects in the mixed litter of F. paniculata and D. glomerata were not significant and therefore not considered any further. We further evaluated the influence of leaf chemistry on litter decomposition, by calculating three mixed models, each including either the C : N ratio, the N content or the phenolic content. All three models included mesocosms as a random factor and location and mesh size as covariates. We evaluated these models for both plant species separately because leaf chemistry was not independent from plant species identity. The significance of each leaf chemistry factor was then evaluated by a Wald chi-square test, using the ‘anova’ function from the ‘car’ r package (Fox & Weisberg 2011).
Table 1. anova table (type III tests) of the C : N ratio (square-root transformed) and the phenolic concentration (log-transformed) in function of leaf state (fresh, senescent, decomposed) and herbivory (presence/absence of grasshoppers)
C : N ratio
Leaf state : Herbivory
Leaf state : Herbivory
Consistent with previous measurements at the site (Gross, Suding & Lavorel 2007), ungrazed F. paniculata fresh leaves had a larger C : N ratio (mean ± SD: 20·2 ± 4·1) than ungrazed D. glomerata fresh leaves (17·5 ± 1·9, Fig. 1). The C : N ratio of D. glomerata leaves was not affected by the presence of grasshoppers, but it was affected by leaf state (Table 1); in particular, decomposed leaves had a lower C : N ratio than fresh and senescent leaves (Fig. 1a). The interaction term between leaf state and herbivory was significant (Table 1), where the C : N ratio of grazed leaves was generally slightly lower than the C : N ratio of ungrazed leaves, except for senescent leaves that had the opposite pattern (Fig. 1a). The pairwise comparisons (Tukey's test) revealed that the C : N ratio of the decomposed leaves in the presence of herbivores was lower than the C : N ratio of both herbivory treatments of the fresh leaves and the senescent leaves with herbivores (Fig. 1a). In contrast, the presence of grasshoppers increased the C : N ratio of F. paniculata leaves overall (Table 1), although pairwise comparisons between grazed and ungrazed leaves of individual leaf states were not significant (Tukey's test, Fig. 1b). Leaf state also affected significantly the C : N ratio of F. paniculata leaves (Table 1). The C : N ratio nearly doubled from 22·1 to 39·6 between fresh and senescent leaves and then decreased in decomposed leaves to 26·2 (Fig. 1b).
The total phenolic concentration in the fresh leaves of both species ranged from 3 to 20 mg g−1. The fresh leaves of D. glomerata contained around twice more phenolic compounds (16·4 ± 2·9 mg g−1) than the fresh leaves of F. paniculata (8·8 ± 1·3 mg g−1, Fig. 2). But as those concentrations decreased more for D. glomerata than for F. paniculata during senescence (Table 2), the senescent leaves of D. glomerata contained less phenolic compounds (4·2 ± 0·4 mg g−1) than the senescent leaves of F. paniculata (5·2 ± 1·3 mg g−1, Fig. 2). During the decomposition process, phenolic compounds continued to disappear in the leaves of F. paniculata, while their concentration increased in D. glomerata (Table 2), so that the decomposed leaves of the two species reached about the same concentrations (about 4·5 mg g−1, Fig. 2). There was a significant interaction between herbivory and leaf state on the phenolic compound concentration of both species (Table 2). Concerning D. glomerata, Tukey's tests indicated that herbivores induced a significant increase (+23%) for fresh leaves. After senescence, less phenolic compounds remained when leaves were grazed, and grazed and ungrazed decomposed leaves contained similar concentrations of phenolic compounds (Fig. 2a). Concerning F. paniculata, herbivory did not seem to induce significant changes in total phenol concentration for fresh leaves or decomposed leaves, but grazed senescent leaves contained significantly less phenolic compounds (−26%) than ungrazed senescent leaves (Fig. 2b).
Table 2. Variation of the concentration of total phenolics during senescence (i.e. from fresh leaves to senescent leaves) and during decomposition (i.e. from senescent leaves to decomposed leaves), for both plant species with and without herbivores. Minus (−) indicates disappearance of phenolics, while plus (+) indicates concentration of phenolics
Total phenolics variation (%)
Based on their UV spectra, 26 flavonoids with an area sufficient to be quantified were detected in the fresh leaves of D. glomerata and 14 in F. paniculata (see Appendix S1 A and B in Supporting Information for HPLC profile examples). Grazing induced a 27% increase in D. glomerata flavonoid content (Fig. 3a, F1,42 = 5·42, P = 0·025), whereas the grazed and ungrazed fresh leaves of F. paniculata contained similar flavonoid amounts (F1,17 = 1·19, P = 0·29). This pattern was confirmed by a multivariate anova (manova) in which each peak was considered as a response variable: in the case of D. glomerata, herbivory had a significant effect (approximate F26,17 = 2·21, P = 0·04), whereas in the case of F. paniculata, it did not (approximate F14,4 = 1·30, P = 0·44). Using t-tests for each peak, we found that seven out of 26 flavonoids of D. glomerata were significantly affected by herbivory (see Table S2, Supporting Information). The concentration of six of these seven flavonoids increased in grazed leaves in comparison with ungrazed leaves, when one decreased. The grazed fresh leaves of D. glomerata contained 66% more flavonoids than the grazed fresh leaves of F. paniculata (Fig. 3a, F1,44 = 25·14, P < 1E-5), and the ungrazed fresh leaves of D. glomerata contained 46% more flavonoids than the grazed fresh leaves of F. paniculata (Fig. 3a, F1,15 = 6·55, P = 0·02).
Eight peaks having a sufficient area to be measured and corresponding to phenolic acids were detected in the fresh leaves of D. glomerata and six in F. paniculata (see Appendix S1 C and D in Supporting Information for chromatogram examples). The major peak (retention time : 31.2 min) was identified as chlorogenic acid, a caffeic acid conjugate. Some minor peaks have been identified as caffeic and ferulic acids, which are derivates of cinnamic acids. The phenolic acid concentrations in the ungrazed fresh leaves of D. glomerata and F. paniculata were not significantly different (Fig. 3b, F1,15 = 0·77, P = 0·39), but in the grazed fresh leaves, they were higher in D. glomerata (F1,54 = 5·57, P = 0·02). The grazed and ungrazed fresh leaves of D. glomerata contained similar phenolic acid amounts (Fig. 3b, F1,44 = 1·30, P = 0·26), as well as the grazed and ungrazed fresh leaves of F. paniculata (F1,25 = 2·94, P = 0·10), a pattern confirmed by multivariate anovas (approximate F8,37 = 1·71, P = 0·13 for D. glomerata and approximate F6,20 = 1·02, P = 0·44 for F. paniculata).
The senescent leaves lost from 10% to 50% of their dry mass during the six winter months (Fig. 4), which is comparable to other measures at this site for the same species (Saccone et al. 2013). The best model with the lowest BIC and AICc included the interaction between herbivores and plant species identity, as well as the interaction between mesh size and location (see Table S3, Supporting Information). Location and mesh size were significant factors (Table S3, Supporting Information), but they did not interact with the plant species and herbivory treatments, so they will not be considered any further. Herbivory by grasshoppers slowed down the decomposition of the F. paniculata litter (−24·5%) and of the mixed litter (−18·1%), but did not significantly affect the decomposition of the D. glomerata litter (+6·4%). Without herbivores, the litter of D. glomerata (D treatment), F. paniculata (F treatment) and the mixed litter (50 treatment) had similar mass loss (around 35% from November to May, the results of the Tukey's tests appear in Fig. 4). In the presence of herbivores, the F. paniculata litter had a lower mass loss than the D. glomerata litter, and the mixed litter had an intermediate mass loss (Fig. 4). The C : N ratio of the F. paniculata senescent leaves had a marginally significant negative effect on their decomposition rate, the N content a marginally significant positive effect and the phenolic content a highly significant positive effect (Table 3). In contrast, none of the D. glomerata chemical traits influenced the decomposition rate (Table 3).
Table 3. Wald chi-square test evaluating the influence of senescent leaf chemistry (C : N ratio, N content and phenolic content) on litter decomposition rate of Festuca paniculata and Dactylis glomerata. ‘d.f.’ means degrees of freedom. When the tests are significant or marginally significant, the sign of the relationship is provided
C : N ratio
C : N ratio
The linkages between herbivory and litter decomposition had almost never been explored for grasses, which contain low amounts of secondary compounds in comparison with, for example, deciduous trees. We found that herbivory by grasshoppers differentially altered the litter decomposition of two functionally contrasted grasses, F. paniculata and D. glomerata (Fig. 4). In order to explain litter decomposition patterns, we explored the effects of herbivory on the biochemical traits of fresh, senescent and decomposed leaves of both plant species.
Unrelated Impacts of Herbivory on Leaf C : N Ratio and Litter Decomposition
Concerning the C : N ratio of leaves, herbivory by grasshoppers did not affect D. glomerata, but it led to a small increase in the C : N ratio of F. paniculata leaves. This increase was not significant when senescent leaves were considered alone (Fig. 1) and significant only when the three leaf states (fresh, senescent and decomposed) were considered simultaneously (Table 1). Festuca paniculata stores carbon and nitrogen below-ground in its perennial tussock, which allows regrowth after mowing (Baptist et al. 2013), and likely also after leaf consumption by herbivores. Regrowing shoots of F. paniculata are poorer in nitrogen (Baptist et al. 2013) because this species is nitrogen limited, which could explain why the grazed leaves had a higher C : N ratio. In contrast with previous findings (Cornwell et al. 2008; Zhang et al. 2008), according to which the C : N ratio and N content are good predictors of litter decomposition, the litter decomposition of the two grass species we studied in this experiment did not significantly depend on these leaf traits (Table 3). However, these previous studies compared species belonging to contrasting functional groups, whereas our tests are intraspecific, with narrow ranges of variation in these traits. The influence of herbivory on litter decomposition was therefore not mediated by the C : N ratio or N content, although in the case of F. paniculata, there was a small nonsignificant trend towards an increase in the C : N ratio and a subsequent decrease in the decomposition rate.
Herbivory Impacts on Litter Decomposition through Changes in Phenolics
Concerning the concentration of phenolic compounds, herbivores had a much greater impact than on the C : N ratio. The grazed fresh leaves of D. glomerata contained much more phenolics than its ungrazed fresh leaves (Fig. 2), which may reflect the expression of defences induced by herbivores. This response seemed to be based upon flavonoids rather than phenolic acids, although both types of compounds have been reported to have antiherbivore properties (Cipollini et al. 2008; Boeckler, Gershenzon & Unsicker 2011). As in other studies (e.g. Norris et al. 2011), we found that phenolics were not a homo-geneous class in terms of biological activity, with phenolic acids acting as constitutive barriers, while flavonoids were actively induced by insect attacks. For D. glomerata, herbivory by grasshoppers induced a significant and specific increase in flavonoid synthesis and accumulation, while the widespread phenolic acids did not markedly change. This strongly suggests that investigations of plant–herbivore interactions should thus go beyond quantification of ‘total phenols’ and incorporate structural determination of the specific molecules and biochemical pathways that are induced by herbivore attacks, because phenolic acids and flavonoids have at least in part distinct metabolic and catabolic pathways. Conversely, in the case of F. paniculata, the grazed and ungrazed leaves contained similar amounts of phenolics, whether flavonoids or phenolic acids. This is probably because F. paniculata mainly relies on constitutive defences against herbivores, such as sclerophylly, silica and higher fibre content. In line with Bryant, Chapin & Klein (1983), (Coley, Bryant & Chapin 1985), we thus found that conservative species (like F. paniculata) tend to have constitutive defences, while exploitative species (like D. glomerata) tend to produce induced defences against herbivores. Plant traits can therefore mediate their responses to herbivory and ultimately the impact of herbivory on litter decomposition (Chapman, Schweitzer & Whitham 2006).
During senescence, phenolics can disappear through different ways, such as biodegradation by epiphytic microbes (Lindow & Brandl 2003), volatilization and leaching (Horner, Gosz & Cates 1988). Most of the phenolic compounds of D. glomerata disappeared during senescence (about 70%, see Table 2), potentially because of its thin cell walls and cuticle, which could favour metabolization by microbes, volatilization and leaching (Tukey 1970). The phenolics disappeared at a higher rate in grazed D. glomerata leaves, probably because they were even more exposed to microbes, volatilization and leaching (Tukey & Morgan 1963). As a result, the grazed and ungrazed senescent leaves of D. glomerata contained similar amounts of phenolics, even if the grazed and ungrazed fresh leaves of these species initially had contrasted phenolic contents. In contrast, only one-third of the phenolic compounds of the ungrazed F. paniculata leaves disappeared during senescence, which may be due to its thicker cell walls and cuticle (Tukey 1970) or to the more recalcitrant nature of F. paniculata metabolites. The phenolics of the grazed leaves disappeared in greater proportions than for the ungrazed leaves (44% vs. 33%, Table 1).Wounding might have led to a greater exposure to microbes, volatilization and leaching of phenolics. As a result, grazed senescent leaves of F. paniculata contained less phenolics than its ungrazed senescent leaves, even if the grazed and ungrazed fresh leaves had similar phenolic contents. However, knowledge on the impact of wounding on the fate of leaf chemicals such as phenols during senescence is lacking.
Although the herbivory treatment was only applied during the leaves’ life, herbivory affected the leaves’ chemistry both during their life and during senescence (Findlay et al. 1996). (i) During the leaves’ life, herbivory increased the phenolic content of the fresh leaves of D. glomerata, but not of F. paniculata. (ii) During senescence, herbivory increased the leaves’ susceptibility to metabolization by microbes, volatilization and leaching. As a result, at the end of senescence, herbivory led to a decrease in the phenolic content of senescent leaves of F. paniculata, but not of D. glomerata. The pattern observed in fresh leaves was thus reversed during senescence due to the physical damage caused by herbivores. Using petiole-galling insects that do not physically alter leaves, Frost et al. (2012) were able to study the effects of chemical induction by herbivores on litter decomposition separately from the physical damage usually caused to leaves by herbivores. However, in the case of grasses and grasshoppers, both effects are usually confounded.
Given the chemical properties of senescent leaves, how can we explain the observed decomposition pattern? The two contrasted pathways through which herbivory affected leaf quality and litter decomposition of D. glomerata and F. paniculata are presented in Fig. 5. The case of D. glomerata is straightforward: the grazed and ungrazed senescent leaves contained similar amounts of phenolics, their C : N ratio was similar, and so was their decomposition rate. In the case of F. paniculata, the senescent grazed leaves contained less phenolics than the ungrazed leaves. Phenolics are known to have contrasted effects on litter decomposition, probably because the broad phenolic group encompasses several structural classes with different chemical properties. For example, Findlay et al. (1996) found that herbivory induced a higher proportion of bound phenolics in poplar leaves, which slowed down decomposition, but Schimel et al. (1996) isolated poplar phenolics and showed that they were used as substrates by microbes and therefore stimulated microbial respiration, while Makkonen et al. (2012) found no clear trend concerning phenolics in a multispecies comparison. We found a positive relationship between the phenolic concentration in senescent leaves and the litter decomposition rate of F. paniculata, but not of D. glomerata, suggesting that the phenolics of F. paniculata can be used as substrates by microbes, as in Schimel et al. (1996). This is a surprising finding because the amount of phenolics in F. paniculata ranges from 4 to 8 mg g−1, which is really low compared to the concentrations observed in the leaves of trees (up to 300 mg g−1, Dudonné et al. 2009) and shrubs (up to 150 mg g−1, Campanella & Bertiller 2011). However, the effects of phenolics on biological processes such as decomposition may depend on their precise chemical structure rather than on their quantity. Our results challenge the view that herbivory generally induces defence compounds that subsequently slow down decomposition (e.g. Findlay et al. 1996). In contrast, we found that in the case of F. paniculata, herbivory decreased the amounts of phenolics in senescent leaves, leading to a slower decomposition.
Lastly, given that herbivory by grasshoppers impacted the litter decomposition rate of F. paniculata and of the mixed litter, herbivory might also have influenced soil parameters such as mineralization rate and nitrogen availability (Belovsky & Slade 2000; Bardgett & Wardle 2003). In order to detect such effects, possibly contrasted depending on the identity of the dominant plant species in the mesocosm, we measured the availability of nitrate and ammonium using ion exchange resin bags (Jaeger et al. 1999), as well as the total nitrogen and carbon contents. The presence of grasshoppers did not affect these soil parameters (details not shown), presumably because the experiment lasted 2 years only.
In summary, herbivory induced the accumulation of phenolics in the fresh leaves of D. glomerata, but most of these phenolics compounds were lost during senescence, so that the decomposition rates of grazed and ungrazed leaves were similar. In contrast, herbivory did not induce any phenolic accumulation in the grazed leaves of F. paniculata, but during senescence, phenolics disappeared in greater proportions in grazed leaves than in ungrazed leaves, due to the probable physical alteration of grazed leaves.
The changes induced by herbivores in fresh leaves must remain in the litter to have an effect on decomposition (after-life effects, Findlay et al. 1996; Chapman et al. 2003). Indeed, the linkage between response traits (here, how fresh leaves are affected by herbivores) and effect traits (here, how does the quality of senescent leaves controls decomposition) is a key to the functioning of ecosystems (Lavorel & Garnier 2002; Lavorel et al. 2013). In the present case, effect and response traits were poorly linked. The processes at play during senescence, in particular, metabolization by microbes, volatilization and leaching, led to a decoupling between the traits of the fresh and senescent leaves, in such a way that the observed decomposition pattern was the opposite of what could have been expected from the fresh leaves quality. We conclude that the combination of chemical and physical modifications of grass leaves due to grazing can explain the diversity of the herbivory effects on their litter decomposition. Hence, even though high herbivore densities were used in this experiment, we believe that herbivory might have complex and contrasted consequences on biogeochemical cycling in grasslands, depending on the dominant plant functional traits in the ecosystem.
We thank Marjorie Bison and Quentin Duparc for their help with the mesocosm experiment, Cindy Arnoldi for help with chemical analyses, Amanda Grand-Veyre for the design of the litter bags, Nicolas Legay for advices during the experiment and two anonymous reviewers for their helpful comments. This research was conducted at the Station Alpine Joseph Fourier (UMS CNRS 3370, France) and on the long-term research site Zone Atelier Alpes (ZAA), a member of the ILTER-Europe network. ZAA publication no. 26.