Parasitic plant litter input: a novel indirect mechanism influencing plant community structure

Authors


Author for correspondence:

Duncan D. Cameron

Tel: +44 (0) 114 222 0066

Email: d.cameron@sheffield.ac.uk

Summary

  • Parasitic plants have major impacts on plant community structure through their direct negative influence on host productivity and competitive ability. However, the possibility that these parasites may also have indirect impacts on community structure (via the mechanism of nutrient-rich litter input) while long hypothesized, has remained unsupported until now.
  • Using the hemiparasite Rhinanthus minor, we established experimental grassland mesocosms to quantify the impacts of Rhinanthus litter and parasitism across two soil fertility levels. We measured the biomass and tissue nutrient concentration of three functional groups within these communities to determine their physiological response to resource abstraction and litter input by the parasite.
  • We show that Rhinanthus alters the biomass and nutrient status of co-occurring plants with contrasting effects on different functional groups via the mechanism of nutrient-rich litter input. Critically, in the case of grass and total community biomass, this partially negates biomass reductions caused directly by parasitism.
  • This demonstrates that the influence of parasitic plant litter on plant community structure can be of equal importance to the much-reported direct impacts of parasitism. We must consider both positive indirect (litter) and negative direct (parasitism) impacts of parasitic plants to understand their role in structuring plant communities.

Introduction

Parasites are potent drivers of ecosystem structure and function, operating both through their direct negative impacts on host performance and by restructuring competitive hierarchies (Press & Phoenix, 2005; Poulin, 2010). In plant communities, such impacts of parasitic plants have been long-recognized (Press & Graves, 1995), where their abstraction of host plant resources (carbon, nutrients and water) and negative impacts on host metabolism (Cameron et al., 2008; Irving & Cameron, 2009) can severely reduce host productivity, altering the competitive balance between host and nonhost plants and leading to a cascade of effects on community structure, diversity, vegetation cycling and zonation (Callaway & Pennings, 1998; Cameron et al., 2009; Irving & Cameron, 2009; Fibich et al., 2010). As well as these direct effects, it has also been hypothesized that parasitic plants may influence community structure through an additional, indirect mechanism via the input of their nutrient-rich litter (Press, 1998). To date, however, this has never been demonstrated.

Parasitic plants produce nutrient-rich litter, as little of their nutrients are resorbed on senescence from their nutrient-rich foliage (Lamont, 1983; Pate, 1995; Quested et al., 2002, 2003a,b, 2005). Because this litter decomposes more rapidly than that of co-occurring nonparasitic species (Quested et al., 2003b; Spasojevic & Suding, 2011) and releases larger amounts of nutrients than litter of other plants within the community (Quested et al., 2002, 2003b), it has been proposed that parasitic plants may provide a benefit to both host and nonhost species by enhancing soil fertility through these nutrient-rich litter inputs (Press, 1998). Indeed, addition of parasitic plant litter may also enhance decomposition of extant litter and thus synergistically increase soil fertility beyond the extent of the direct nutrient input (Watson & Herring, 2010). For example, the hemiparasite Bartsia alpina has been estimated to increase soil nitrogen (N) inputs to a sub-arctic heathland by > 50%, and pots studies using individual phytometer seedlings have shown that B. alpina litter (albeit in considerable quantities) can increase the productivity of co-occurring species by up to 51% compared to litters of other co-occurring species (Quested et al., 2003b). Indirect measures have also indicated nutrient enrichment (as suggested by dilution of a 15N tracer pool) in semi-natural grasslands infected with the hemiparasites Rhinanthus angustifolius and R. minor (Ameloot et al., 2008). In managed semi-natural grasslands the litter of annual hemiparasites may be of even greater significance because, due to their early senescence, they generate litter that can contribute to nutrient cycling while much of the biomass of their co-occurring species is removed at hay-cut. Despite its apparent capacity to increase nutrient supply and enhance growth of other plants, the effects of parasitic plant litter on community structure have never been determined (Quested et al., 2003b). Given such a positive effect could clearly act to oppose the negative effects of parasitism, understanding the impacts of parasite litter on community structure, and the strength of this mechanism compared to the impacts of parasitism, is essential if we are to fully understand how parasitic plants influence plant community structure and function. In addition, the introduction of nutrients through parasitic plant litter is likely to influence the structure soil decomposer communities such as fungi, bacteria or arthropods (Watson, 2009). The effect of parasite litter should depend largely on the equality with which this nutrient resource benefits co-occurring species and functional groups. If all the components of a community benefit equally from the hemiparasite-litter derived nutrients then there may be no change in community composition, other than an increase in net productivity. However, this is unlikely because the relationship between nutrient supply and plant performance is rarely equal across all species. Therefore, it is likely that some species or groups of species will receive a greater benefit from the litter-derived nutrient input, and that the prominence of these groups may increase at the expense of others. Indeed, it has been proposed that some parasitic plants may redistribute mineral resources from fast-growing nutrient rich hosts to slower growing, less abundant members of the plant community (Press, 1998), but it is also possible that the nutrients in the hemiparasite litter could predominantly benefit the hosts from which they were initially derived. Which functional groups benefit from litter fertilization may depend in part on the functional traits of that group. For instance, in grasslands (as modelled in our study), the shallow and highly branching fine root systems of grasses may have better access to parasite litter input compared to forbs that may benefit less given that they often have simpler, more mycorrhiza-dependent, tap root systems (Levang-Brilz & Biondini, 2003).

Rhinanthus minor is an annual root hemiparasite and a common constituent of grassland communities across temperate northern Europe and North America (Westbury, 2004). Crucially for the community level impacts of R. minor, the damage that this parasite inflicts on different host species varies, (Cameron et al., 2006), with grasses generally showing the greatest reduction in biomass while forbs (nonleguminous dicots) exhibit the least damage and legumes show variable responses (Ameloot et al., 2005). This differential resistance, and the differential host damage that results from it, enables R. minor to shift the competitive balance between functional groups and hence alter grassland community structure by enhancing forb abundance at the expense of grasses (Gibson & Watkinson, 1991, 1992; Pywell et al., 2004; Cameron et al., 2006). Although we have a strong understanding of the physiological mechanisms underpinning host damage and variable resistance, as well as an understanding of how these impacts influence the interactions between host and nonhost species, the way in which such effects might be ameliorated by the indirect impact of nutrient-rich parasite litter inputs remains unknown (Press, 1998; Press & Phoenix, 2005). Rhinanthus minor senesces early in the season so its litter may represent a particularly important nutrient input in managed grasslands as much of the biomass of the co-occurring perennial species is removed at hay-cut. Furthermore, both the direct effect of parasitism and the indirect effect of nutrient-rich litter are likely to be modulated by soil nutrient concentrations. Indeed, a recent pot study has suggested that R. minor may only have a detrimental effect on its hosts under high fertility conditions (Cameron et al., 2005). By contrast, it could be expected that R. minor litter will have a greater impact on community structure in less fertile conditions where nutrients are more limiting. Given parasites such as R. minor occur in grasslands from low to intermediate fertility, it is also important to understand the role soil fertility has on mediating the impacts and interactions between parasitism and litter inputs.

Here we describe the findings from a unique, long-term experiment designed to investigate the impact of nutrient-rich hemiparasite litter and parasitism on the structure of grassland communities. Using model ‘mesocosm’ experimental communities of a meadow grassland, with factorial treatments of R. minor parasitism and litter inputs, duplicated at high (fertilized) and low (natural) soil nutrient concentrations, we tested the following hypotheses: parasite litter will benefit grasses more than forbs and legumes, resulting in increased tissue nutrients and biomass of grasses that will be less (or not) apparent in forbs and legumes; R. minor parasitism will have a significant impact on grassland community structure, decreasing the abundance of grasses and increasing that of forbs with legumes showing an intermediate response; tissue nutrient and biomass changes caused by parasitism will be opposite to, and therefore negated by, parasite litter inputs; the impact of R. minor litter will be greater under less fertile soil conditions while the impact of R. minor parasitism will be greater under higher soil fertility.

Materials and Methods

Experimental design

Model grassland communities were established in 2005 outside at the University of Sheffield (UK) experimental gardens. Each mesocosm consisted of a polypropylene box (35 cm × 35 cm × 12 cm deep) filled to a depth of 12 cm with homogenized Rendzina soil from a traditionally managed calcareous hay meadow at Lathkill Dale, Derbyshire, UK (GPS ref: N 53 11.539, W 1 45.767). The soil was collected from the top 20 cm and sieved through a 1-cm mesh before use. Two species of grass (Anthoxanthum odoratum L. and Festuca ovina L.); two forb species (Primula veris L. and Sanguisorba minor Scop.) and two species of legume (Lotus corniculatus L. and Trifolium pratense L.), all representative of common calcareous grassland species, were collected as mature plants and transplanted into the mesocosms. Plant communities were allowed to establish for a year before treatment applications.

In 2006, the mesocosm treatments were established in a fully factorial semi-randomized design. Nutrient treatments were assigned to a random half of each block of eight mesocosms, with each half receiving either no fertilizer or annual applications of 11.9 g m−2 of Yara ‘New 52’ 21 : 8: 11 NPK fertilizer per mesocosm. This fertilizer treatment is equivalent to 100 kg N ha−1 which is comparable to the fertilizer treatments of some previous studies (Smith et al., 2000, 2008). Nutrient treatments were randomly allocated to one half of each block rather than individually allocated to each mesocosm to reduce the likelihood of nutrient transfer between mesocosms. Within each nutrient concentration, mesocosms received either 3 g of Rhianthus minor L. seed each year or no R. minor seed and either R. minor litter or no R. minor litter. Each treatment combination was therefore replicated eight times to give a total of 64 mesocosms within eight blocks. Mesocosms receiving no R. minor seed were weeded in early spring each year to remove emerging parasites (if any) before they could become established. In 2008 the mean R. minor biomass in each infected mesocosm was 9.6 g (95% CI: 8.4, 10.8 g) and was not significantly different between treatments (Supporting Information Table S1, Fig. S1).

Rhinanthus minor litter was generated by harvesting the mature parasite from all R. minor + mesocosms at the first signs of senescence. The harvested R. minor was then bulked and homogenized in two separate batches according to the nutrient treatment of the mesocosms from which it was derived. This material was then divided between litter and R. minor + litter treatments within each nutrient concentration, so that litter derived from the high nutrient treatment was only returned to the high nutrient mesocosms and litter from unfertilized mesocosms was only returned to unfertilized mesocosms. For example, in 2007 a total of 156 g of R. minor was harvested from the unfertilized mesocosms and 146 g was harvested from the fertilized mesocosms. Both of these masses of litter were divided between 16 mesocosms (eight litter treatments and 8 R. minor + litter treatments), so that each unfertilized litter + mesocosm received 9.8 g litter and each fertilized litter + mesocosm received 9.1 g litter. The overall mean tissue N and phosphorus (P) concentrations were 17.0 mg g−1 (95% CI: 15.8, 18.1) and 3.18 mg g−1 (95% CI: 3.99, 3.55), respectively, and did not differ significantly between the fertilized and unfertilized treatments (Table S2). However, because of the slightly different masses of litter added, total nutrient inputs did differ between the nutrient treatments, with 146.5 mg (95% CI: 131.4, 162.6) and 175.4 mg (95% CI: 162.1, 183.2) of N being added to the fertilized and unfertilized treatments, respectively (Table S2). The difference in the total P added through litter to each mesocosm was not significantly different, with 27.1 mg (95% CI: 25.2, 29.0) mg being added to the fertilized communities and 33.0 mg (95% CI: 30.3, 39.0) to the unfertilized communities (Table S2).

We did not attempt to collect freshly fallen litter since R. minor retains most of its nutrients within its leaves during senescence, so live leaves and litter have very similar nutrient concentrations (Quested et al., 2002). Also, collection of fallen litter is practically impossible in a dense grassland sward. Mesocosms were managed by clipping to 3 cm above the soil level each year in late summer after R. minor had been removed, replicating the management of a traditional hay meadow.

Biomass and tissue nutrient determination

In October 2008, after R. minor had been removed for litter production, the host community above-ground biomass was harvested to 3 cm above the soil surface from each mesocosm and sorted to functional group level (separated into grasses, forbs (excluding R. minor) and legumes). The harvested material was oven dried (80°C, 48 h), before weighing, then homogenised using an analytical grinder and a subsample was analysed for tissue N and P concentration by flow injection analysis (FIAflow2; Burkard Scientific, Uxbridge, UK) of Kjeldahl digested samples (Allen, 1989). Five subsamples of homogenised R. minor litter from both the fertilized and unfertilized communities was similarly processed and analysed for tissue N and P concentration.

Statistical analyses

The main effects of R. minor parasitism, litter application and soil fertility on biomass and tissue nutrient concentration were determined for each functional group using a three-way ANOVA with two separate error strata. The R. minor and litter factors were nested with the nutrient factor which was in turn nested within the block factor. This nested analysis was required because the nutrient treatment was randomly allocated to one half of each block of eight mesocosms, and within each half the Rhinanthus and litter treatments were allocated randomly. Equivalence of variance was confirmed by visual inspection of residual vs fitted values. Effect sizes are reported throughout as percentage changes, because these are more intuitive for biomass or nutrient-concentration differences than standard effect sizes, and are shown with 95% bias-corrected accelerated percentile bootstrapped confidence intervals. Percentage changes were calculated assuming that interaction terms were not significant. R. minor litter subsamples from the fertilized and unfertilized communities were compared with a two sample t-test assuming unequal variance. All analysis was carried out using the statistical package R (R Core Development Team, 2011).

Results

Aboveground biomass

Across all treatments, the addition of Rhinanthus minor litter caused a significant 10.0% increase in total community biomass (Table 1, Fig. 1d). By contrast, parasitism by R. minor caused an overall 25.8% decrease in total community biomass; however, there was a significant interaction between R. minor parasitism and soil fertility, with R. minor causing a greater reduction in total biomass in unfertilized communities than in fertilized communities (Table 1, Fig. 1d).

Table 1. Nested ANOVA results for the impact of soil fertility, Rhinanthus minor litter and parasitism on grass, forb (excluding R. minor), legume and total community above ground biomass
Functional groupSource of variationdf F P Percentage change (95% CI)
  1. Percentage change in biomass is shown for each main effect with 95% bias-corrected accelerated percentile bootstrapped confidence intervals in brackets.

Total community Nutrients 1,7 17.64 0.004 18.0 (10.4, 26.1)
Rhinanthus 1,42 101.2 < 0.001 −25.8 (−30.6, −20.4)
Litter 1,42 10.66 0.002 10.0 (1.2, 18.2)
Rhinanthus × Litter1,420.0030.955 
Rhinanthus × Nutrients 1,42 5.010 0.031  
Litter × Nutrients1,420.5150.477 
Rhinanthus × Nutrients × Litter1,420.0340.854 
Grasses Nutrients 1,7 7.452 0.030 15.6 (5.0, 26.3)
Rhinanthus 1,42 84.80 < 0.001 −27.9 (−33.1, −21.9)
Litter 1,42 7.377 0.010 10.0 (0.2, 19.5)
Rhinanthus × Litter1,420.0450.892 
Rhinanthus × Nutrients1,423.7900.058 
Litter × Nutrients1,420.3460.559 
Rhinanthus × Nutrients × Litter1,420.1900.665 
ForbsNutrients1,70.0000.996−0.3 (−17.8, 24.9)
Rhinanthus 1,420.6130.4389.5 (−4.3, 26.9)
Litter1,422.0710.157−15.4 (−30.0, 2.2)
Rhinanthus × Litter1,422.0820.156 
Rhinanthus × Nutrients1,423.1300.084 
Litter × Nutrients1,420.6960.408 
Rhinanthus × Nutrients × Litter1,423.6990.061 
Legumes Nutrients 1,7 16.18 0.005 111.8 (44.5, 212.7)
Rhinanthus 1,42 5.133 0.029 −23.5 (−40.6, −1.1)
Litter 1,42 7.706 0.008 39.0 (9.1, 70.87)
Rhinanthus × Litter1,420.3460.560 
Rhinanthus × Nutrients1,420.5220.474 
Litter × Nutrients1,421.1540.289 
Rhinanthus × Nutrients × Litter1,420.5860.448 
Figure 1.

Total aboveground biomass of (a) grasses, (b) forbs (excluding Rhinanthus minor), (c) legumes, (d) total community (total biomass is that of all plants except R. minor), for all eight treatment combinations. Box plots represent median (midline), quartiles (box), maximum and minimum (whiskers) biomass with outliers represented by points.

The biomass of each functional group showed contrasting responses to parasitism by R. minor and litter application. The addition of R. minor litter increased grass biomass overall by 10.0%, while R. minor parasitism reduced grass biomass by 27.9% (Table 1, Fig. 1a). The main effect of soil fertility significantly increased grass biomass by 15.6%; however, grass biomass was subject to a marginally significant interaction between R. minor parasitism and soil fertility level which indicated that R. minor caused a greater proportional reduction in grass biomass at lower rather than higher soil fertility (Table 1, Fig. 1a).

Forb biomass showed only a marginally significant three-way interaction between soil fertility, litter application and R. minor infection (Table 1): at low fertility neither litter nor R. minor parasitism affect forb biomass; however, with higher soil fertility litter reduced forb biomass, but only when R. minor was absent (Fig. 1b).

Litter application increased legume biomass by 39.0% across treatments, while R. minor parasitism decreased legume biomass by 23.5% (Table 1, Fig. 1c). Legume biomass was 111.8% greater under high soil fertility compared to low soil fertility (Table 1, Fig. 1c).

Plant nutrient impacts

Rhinanthus minor parasitism increased tissue N and P concentrations of grasses (Tables 2, 3, Fig. 2a) by 14.9% and 20.1%, respectively, while reducing total aboveground grass N and P pools (Tables S3–S6, Figs S2, S3). However, litter and soil fertility had no significant impact on grass tissue N and P concentration (Tables S3–S6, Figs S2, S3).

Table 2. ANOVA results for the impact of soil fertility, Rhinanthus minor litter and parasitism on grass, forb (excluding R. minor) and legume tissue nitrogen (N) concentration in fertilized and unfertilized communities
Functional groupSource of variationdf F P Percentage change (95% CI)
  1. Percentage change in tissue N concentration is shown for each main effect with 95% bias-corrected accelerated percentile bootstrapped confidence intervals in brackets.

GrassesNutrients1,75.07420.059−4.8 (−7.6, 2.5)
Rhinanthus 1,42 15.82 < 0.001 14.9 (5.9, 23.8)
Litter1,420.5610.457−2.6 (−10.7, 6.3)
Rhinanthus × Litter1,421.4170.240 
Rhinanthus × Nutrients1,422.1450.150 
Litter × Nutrients1,420.1910.664 
Rhinanthus × Nutrients × Litter1,420.0450.833 
ForbsNutrients1,75.1610.057−9.0 (−16.5, 3.8)
Rhinanthus 1,42 26.00 < 0.001 35.3 (21.9, 47.0)
Litter1,420.5380.4674.4 (−8.1, 17.1)
Rhinanthus × Litter1,420.8940.350 
Rhinanthus × Nutrients1,422.4530.125 
Litter × Nutrients1,420.0240.878 
Rhinanthus × Nutrients × Litter 1,42 5.679 0.022  
LegumesNutrients1,70.04460.8399.7 (−14.2, 48.2)
Rhinanthus 1,420.1910.6654.3 (−9.8, 21.2)
Litter1,420.5380.468−6.8 (−19.2, 7.0)
Rhinanthus × Litter1,420.5740.453 
Rhinanthus × Nutrients1,421.0670.308 
Litter × Nutrients 1,42 7.805 0.008  
Rhinanthus × Nutrients × Litter1,420.0720.789 
Table 3. ANOVA results for the impact of soil fertility, Rhinanthus minor litter and parasitism on grass, forb (excluding R. minor) and legume tissue phosphorus (P) concentration in fertilized and unfertilized communities
Functional groupSource of variationdf F P Percentage change (95% CI)
  1. Percentage change in tissue P concentration is shown for each main effect with 95% bias-corrected accelerated percentile bootstrapped confidence intervals in brackets. Data pertaining to this statistical summary can be found in Fig. S4.

GrassesNutrients1,70.0090.926−1.1 (−8.5, 18.2)
Rhinanthus 1,42 7.851 0.008 20.1 (9.3, 40.8)
Litter1,422.6690.110−10.1 (−16.4, 0.52)
Rhinanthus × Litter1,421.2990.261 
Rhinanthus × Nutrients1,420.2130.647 
Litter × Nutrients1,420.6880.411 
Rhinanthus × Nutrients × Litter1,420.8980.349 
ForbsNutrients1,70.8500.387−4.4 (−14.6, 18.4)
Rhinanthus 1,421.6740.20315.3 (1.92, 32.2)
Litter1,421.7710.19115.8 (−.12, 37.3)
Rhinanthus × Litter1,420.6420.428 
Rhinanthus × Nutrients1,420.8890.351 
Litter × Nutrients1,420.3140.578 
Rhinanthus × Nutrients × Litter1,421.5660.218 
LegumesNutrients1,70.8320.39228.9 (−5.7, 93.0)
Rhinanthus 1,420.9600.33314.3 (−6.44, 38.99)
Litter1,420.5030.482−9.2 (−24.8, 11.4)
Rhinanthus × Litter1,421.0800.305 
Rhinanthus × Nutrients1,420.5280.471 
Litter × Nutrients1,423.3020.076 
Rhinanthus × Nutrients × Litter1,420.1180.733 
Figure 2.

Tissue nitrogen concentration of (a) grasses, (b) forbs (excluding Rhinanthus minor), (c) legumes, for all eight treatment combinations. Box plots represent median (midline), quartiles (box), maximum and minimum (whiskers) biomass with outliers represented by points.

Forb tissue N concentration showed a significant three-way interaction between R. minor parasitism, soil fertility and litter (Table 2). At high soil fertility, the addition of litter in the absence of R. minor caused a slight reduction in forb tissue N concentration, whereas the addition of litter in the presence of R. minor caused an increase in forb tissue N (Fig. 2b). Conversely, at low soil fertility the addition of litter and the presence of R. minor had little impact of forb tissue N concentration (Fig. 2b). None of the treatments had a significant effect on forb tissue P concentration (Table 2).

Legume tissue N showed a significant interaction between litter application and nutrient concentration (Table 2). Under high soil fertility the addition of litter caused a reduction in legume tissue N concentration whereas under low soil fertility the addition of litter caused an increase in legume tissue N (Fig. 2c). None of the treatments had a significant impact on legume tissue P concentration (Table 3, Fig. S4). Fertilizer application increased the total aboveground N and P pool while R. minor infection decreased the legume aboveground N pool and caused a marginal reduction in the legume aboveground P pool (Tables S3, S4, Figs S2, S3).

Fertilizer application increased the total community aboveground N and P pools while R. minor infection caused a decrease in total aboveground N and P. When considered in the absence of R. minor, litter caused an increase in the total community aboveground N and P pools (Tables S3, S4, Figs S2, S3).

Discussion

The direct effects of parasitic plants on the structure and function of host communities has been investigated in detail (Irving & Cameron, 2009), however, although it has been hypothesized for more than a decade that parasitic plants have the potential to exert significant, indirect effects on plant community composition through their nutrient-rich litter (Press, 1998), this has never previously been demonstrated. Here, for the first time, we show that hemiparasite litter can alter plant community structure and that such impacts can, in some cases, be of a similar magnitude to the direct impact of parasitism. Given that the effects of litter are often in the opposite direction to those of parasitism, this study has also shown that, intriguingly, litter inputs can in fact partially negate the impact of parasitism.

Impacts of R. minor litter on plant community structure

Rhinanthus minor litter increased the total biomass of both fertilized and unfertilized communities. Such benefits for productivity resulting from nutrient-rich parasite litter have been observed before in the hemiparasite Bartsia alpina, which stimulated growth of a grass and a dwarf shrub in pot studies (Quested et al., 2003b), and in mistletoe (Amyema miquelii) that increased productivity of understorey plants (March & Watson, 2007). However, our study is the first to show that the impacts of parasite litter on biomass differ between the component plant functional groups within the community and can hence drive changes in community structure.

The R. minor in our communities produced litter with nutrient concentrations similar to those seen in the live tissues of the species in our communities and in excess of those measured in litter produced by nonparasitic grassland species in other studies (Bloemhof & Berendse, 1995; Aerts et al., 2003; Koukoura et al., 2003). This early-senescing, nutrient-rich parasite litter is particularly influential in semi-natural grassland because much of the other biomass is removed during hay-cut before falling as litter. It was hypothesized that grasses would benefit most from R. minor litter because their shallow, fine, highly branched root systems would be best able to acquire nutrients from parasite litter decomposing at the soil surface – a benefit that would be less clear in forbs and legumes with their simpler, often more mycorrhiza-dependent root systems (Levang-Brilz & Biondini, 2003). This was partially supported by our finding whereby litter increased the productivity of grasses (by 10%) while forbs showed little clear benefit. Although the benefit of litter nutrients was not apparent in grass N and P concentrations (which changed little in response to litter input), this may simply be due to a growth dilution effect of the greater productivity grasses under the litter treatment.

Crucially, the addition of R. minor litter to R. minor-parasitized grasses increased their biomass, restoring their productivity towards that of unparasitized grasses. This shows that realistic litter inputs from R. minor can at least partially negate the impacts of its parasitism.

Somewhat surprisingly though, legumes showed by far the greatest biomass increase (39.0%) with R. minor litter addition. In fact, with stronger competition from the increase in grass biomass, we may have expected a decline in legumes. However, productivity in such meadow grasslands on rendzina soil can be P limited (Phoenix et al., 2003). Any increase in P supply may be disproportionately beneficial to an N fixing plant because they already have access to a sufficient supply of N. This would also be consistent with the decrease in legume N concentration (through growth dilution), as seen in the high soil-fertility mesocosms. However, in our low soil-fertility mesocosms, litter application increased legume tissue N concentration. This difference may be due to different impacts of fertility on legume biomass. In the low soil-nutrient communities the productivity of the legumes was lower so they may have been limited by light availability due to shading from taller grasses, leading them to concentrate N in their tissues. Also, the ratio of litter-derived nutrients to final legume biomass would have been greater in the low soil-nutrient communities (due to reduced legume biomass under this treatment) and this may have further contributed to the concentration of N in legume tissues. However, it is also possible that this may be a result of the higher absolute mass of N that was added through litter to the low-nutrient mesocosms.

While forbs showed little response to R. minor litter in terms of biomass production, there was a significant three-way interaction determining forb tissue N across all communities. In the low soil-fertility communities R. minor litter and parasitism did not appear to have a major impact on forb tissue N. However, in the fertilized communities the addition of litter caused a decrease in forb tissue N in the absence of R. minor but an increase in forb tissue N in the presence of R. minor. This is the same as the pattern of response seen in forb biomass in the fertilized communities, and both trends can probably be explained by considering competition between forbs and grasses. Generally it is expected that grasses will dominate forbs within a plant community (del-Val & Crawley, 2005; Cameron et al., 2009), but where R. minor parasitism suppresses grasses, competition with forbs for nutrients is reduced, hence the increase in forb foliar N concentration. The fact that such reduction in competition for nutrients does not increase forb biomass suggests they are still outcompeted for light by the taller statured grasses. Quite why litter might reduce forb N when R. minor is absent is not clear, but caution should be taken in interpreting that effect as the reduction is small in magnitude and three-way interactions can be difficult to interpret without further experimentation.

Such understanding of parasitic plant litter impacts on plant community structure is important not only in terms of our ecological understanding, but also for grassland conservation and restoration. In particular, given that Rhinanthus species are used to aid restoration of species-rich grasslands from improved pastures through their promotion of forbs at the expense of grasses (Davies et al., 1997; Pywell et al., 2004), the fact that litter inputs appear to benefit grasses suggests that the timing of hay-cut (and hence the quantities of Rhinanthus litter allowed to enter the system) need also be taken into consideration. Furthermore, as it has already been show that the direct impacts of parasitic plants extend beyond plant communities to other tropic levels both above- and belowground (Bardgett et al., 2006; Ewald et al., 2011; Watson et al., 2011), we might also expect the indirect litter impacts to have similarly multitrophic effects.

Impacts of R. minor parasitism on plant community structure

In common with other studies (Ameloot et al., 2005), R. minor parasitism reduced the total aboveground biomass of the co-occurring plant community, with this change being primarily driven by reductions in the biomass of grasses. Grasses have been shown to be highly susceptible to R. minor because they are unable to mount a successful resistance response to the parasite (Cameron & Seel, 2007). This could also explain the increase in grass N and P concentrations in communities with R. minor parasitism, because reduced growth could result in a concentrating effect on host nutrients, even though R. minor caused a decrease in total grass aboveground N and P stocks (Cameron et al., 2008). However, the increase in grass tissue N and P concentration could also be the result of an increased root : shoot ratio, which is known to be a consequence of R. minor infection (Jiang et al., 2004).

Infection by R. minor had a substantial negative impact on legumes, reducing their biomass by 23.5%. This is perhaps unsurprising as legumes are known to be high-quality hosts for R. minor (Jiang et al., 2008). We might have expected the negative impact of the hemiparasite on legume biomass to be ameliorated by decreased competition from the parasitized grasses. However, in this study the proportional decrease in biomass was similar in both grasses and legumes. Surprisingly, R. minor parasitism did not result in any clear increase in forb biomass. Although there is a general trend apparent in the literature for R. minor infection to increase the biomass of forbs and decrease that of grasses, there is considerable variability in its magnitude. Various factors may underpin this variability, but our work suggests that soil fertility is likely to be important in determining the community-level impact of R. minor infection (see later). It is also possible that individual species may respond differently to R. minor infection, with some forb species being promoted more strongly than others, and that in our experimental communities we have inadvertently chosen species which respond less strongly to parasite-induced reductions in grass competitive ability. However, the increase in forb biomass is likely to be a long-term, more indirect effect of declining grass biomass and it is possible that the 3 yr duration of this experiment was not sufficient to detect a significant change in forb abundance.

Host community responses in fertilized and unfertilized soils

Contrary to our hypothesis, litter impacts were similar across fertilized and unfertilized communities. However, the relatively small magnitude of community biomass response to nutrient addition compared with that we could expect for intensively managed grassland suggests that this manipulation was relatively modest and therefore would not have diminished the importance of litter nutrients in the high soil-fertility mesocosms. Our hypothesis that the impacts of parasitism would be greatest in the fertilized microcosms was also seen not to be true. Indeed, R. minor caused a greater reduction in biomass under low rather than high fertility. This was largely a consequence of the reduction in grass biomass, which showed a similar trend to total community biomass and a similar (although marginally significant) interaction between soil fertility and parasitism. This suggests that the fertilized communities were more able to compensate for the detrimental effects of the parasite, possibly by increasing nutrient uptake from the soil. Previous pot studies have also shown that increased nutrient inputs may alleviate the impact of hemiparasitic plants on their hosts (Cechin & Press, 1994); however, our study contrasts with another which suggested that R. minor only has a negative impact on its hosts under high-fertility conditions (Cameron et al., 2005). This may be because pot studies do not allow the complex interactions between parasitism and competition to arise as they have done in our more natural and diverse mesocosms communities.

Litter and parasitism impacts on plant communities

In the field, R. minor has a dynamic and sometimes patchy distribution. Through its impact on nutrient cycling via nutrient-rich litter and the alteration of competitive interactions, R. minor has the potential to greatly increase habitat heterogeneity and so potentially also species richness at the field level. Indeed, the spatial scales over which the direct effects of parasite resource abstraction and the indirect effects of parasite litter influence surrounding plants is likely to differ because the sphere of influence of the parasite's roots, and hence haustoria, may be smaller than that of the litter which can be spread further in the meadow (e.g. by wind or during hay-cut). Consequently, patches within the community have the potential to be influenced by the indirect effects of litter without the influence of the direct effects of parasitism (and vice versa), which may further promote community heterogeneity.

This study has highlighted that hemiparasitic plants can have both positive and negative direct and indirect impacts on both host and nonhost species. It reveals for the first time that parasitic plant litter can influence plant community structure, and that this effect can be opposite to, and therefore partially negate, the impacts of parasitism.

Acknowledgements

We would like to thank Alex Williams and Tom Parker (University of Sheffield) for assistance in the field and Irene Johnson, Sarah Hawker and Dr Jill Edmondson (University of Sheffield, UK) for expert technical support. This work was funded by the NERC through grants NE/C510124/1 (to M.C.P. & G.K.P.) and NE/E014070/1 (to D.D.C.) and a NERC doctoral training grant NE/G524136/1 (to J.P.F., G.K.P., M.C.P. and D.D.C.). D.D.C. is supported by a Royal Society University Research Fellowship (UF090328).

Ancillary