Closely related parasitic plants have similar host requirements and related effects on hosts

Abstract The performance of root hemiparasites depends strongly on host species identity, but it remains unknown whether there exist general patterns in the quality of species as hosts for hemiparasites and in their sensitivity to parasitism. In a comparative approach, the model root hemiparasites Rhinanthus minor and R. alectorolophus were grown with 25 host species (grasses, forbs, and legumes) at two nutrient levels. Hosts grown without parasites served as a control. Host species identity strongly influenced parasite biomass and other traits, and both parasites grew better with legumes and grasses than with forbs. The biomass of R. alectorolophus was much higher than that of R. minor with all host plants and R. alectorolophus responded much more strongly to higher nutrient availability than R. minor. The performance of the two species of Rhinanthus with individual hosts was strongly correlated, and it was also correlated with that of R. alectorolophus and the related Odontites vulgaris in previous experiments with many of the same hosts, but only weakly with that of the less closely related Melampyrum arvense. The negative effect of R. minor on host biomass was less strong than that of R. alectorolophus, but stronger relative to its own biomass, suggesting that it is more parasitic. The impact of the two parasites on individual hosts did not depend on nutrient level and was correlated. Several legumes and grasses were tolerant of parasitism. While R. minor slightly reduced mean overall productivity, R. alectorolophus increased it with several species, indicating that the loss of host biomass was more than compensated by that of the parasite. The results show that closely related parasites have similar host requirements and correlated negative effects on individual hosts, but that there are also specific interactions between pairs of parasitic plants and their hosts.


| INTRODUC TI ON
Root-hemiparasitic plants have green leaves and are photosynthetically active, but attack the roots of other plants and extract water and solutes from them (Cameron & Phoenix, 2013;Phoenix & Press, 2005). While hemiparasites may use a wide range of plant species as their hosts, different species vary in their quality as hosts and the identity of the host strongly influences the growth of hemiparasites (Calladine et al., 2000;Hautier et al., 2010;Matthies, 2017;Nge et al., 2019;Rowntree et al., 2014;Sandner & Matthies, 2018).
The suitability of a species as a host for a hemiparasite depends on the quantity and quality of compounds the parasites obtains from the host (Atsatt, 1983;Govier et al., 1967) and on the strength of resistance it has against the attack by the hemiparasite (Cameron et al., 2006). Because hemiparasites usually have a very reduced root system (Matthies, 2017), the uptake of water and nutrients from the host is thought to be the most important benefit of parasitism, but hemiparasites can also obtain significant amounts of carbon from their hosts (Press et al., 1988;Tennakoon & Pate, 1996;Těšitel, Plavcová et al., 2010;Těšitel, Těšitelová, et al., 2015).
Based on studies with species of the genus Rhinanthus, it has been concluded that legumes are particularly good hosts for hemiparasites, followed by grasses, while nonleguminous forbs are less suitable as hosts (Cameron & Phoenix, 2013). However, in a large study with the hemiparasite Melampyrum arvense forbs were on average better hosts than grasses (Matthies, 2017). There is also considerable variation in host quality within functional groups (Hautier et al., 2010;Matthies, 2017;Rowntree et al., 2014). For example, in spite of their high N-content not all legumes are good hosts for hemiparasites (Matthies, 1998;Nge et al., 2019;Radomiljac, 1999).
The legume Anthyllis vulneraria was found to be a poor host for both Melampyrum and Rhinanthus, and Onobrychis viciaefolia was an unsuitable host for R. alectorolophus (Matthies, 2017;Sandner & Matthies, 2018). Moreover, the results of studies on the quality of individual species as hosts for hemiparasites have often been inconsistent. For example, Anthyllis was a good host for Euphrasia ssp. (Yeo, 1964), and while Trifolium repens was a good host for Rhinanthus angustifolius (De Hullu, 1984), it was a poor host for two species of Odontites (Snogerup, 1982). These inconsistent results could be due to specific interactions between hemiparasite-host pairs, but could also be due to differences in experimental conditions. Because in most studies, a hemiparasite species was grown with only one or a few host species, effects of host species and experimental conditions cannot be separated, which severely restricts the value of comparisons of the relative performance of parasites with the various hosts. It is thus not known whether the performance of different species of hemiparasites with individual host species is correlated.
Hemiparasites often have strong negative effects on the growth of their host plants, because they extract water, nutrients, and carbon from them and may reduce host photosynthesis (Matthies, 1995;Phoenix & Press, 2005). However, there is strong variation in the sensitivity of plant species to parasitism. While growth and reproduction of some species are strongly reduced by a hemiparasite, others are resistant against parasite attack (Cameron et al., 2006), and some species are tolerant of parasitism; that is, they are good hosts and provide strong benefits to the parasites but are not harmed by them (Matthies, 2017;Sandner & Matthies, 2018). However, it is not known whether the resistance or tolerance of an individual host species is parasite-specific, or whether the response of potential host plants to different species of parasites is similar.
In most experiments, root hemiparasites have reduced overall productivity (Ameloot et al., 2005;Hautier et al., 2010;Matthies, 1995Matthies, , 1996, indicating that their resource use efficiency is lower than that of their host plants (Matthies, 1995; F I G U R E 1 (a) Rhinanthus minor and (b) Rhinanthus alectorolophus (a) (b) Těšitelová, et al., 2015). It has therefore been proposed to use hemiparasites to reduce the productivity of grasslands of interest for conservation (Demey et al., 2015;Pywell et al., 2004;Těšitel et al., 2017;Westbury et al., 2006). However, although it has been predicted that the productivity of hemiparasite-host systems will always be lower than that of the communities without the parasite (Hautier et al., 2010), in some studies hemiparasites had no effect on overall productivity or even increased it (Joshi et al., 2000;Sandner & Matthies, 2018).
The aim of the present study was to comparatively investigate the interactions of two closely related hemiparasites, Rhinanthus alectorolophus (Scop.) Poll. and R. minor L. (Orobanchaceae; Figure 1), with a large number of species. Rhinanthus species have been widely used as models for the investigation of hemiparasite-host relationships (Cameron & Phoenix, 2013;Matthies, 1995;Rowntree et al., 2014;Sandner & Matthies, 2018;Seel & Press, 1994;Těšitel, Těšitelová, et al., 2015). The two species of Rhinanthus were grown with 25 different host species, many of which had also been used in some previous experiments with hemiparasites. This made it possible to compare the performance of the two parasites with the individual hosts and with that of other hemiparasites grown with the hosts in previous studies. The host species were in addition grown without a parasite to compare the effects of the two parasites on the host species. As hemiparasite-host relations may be influenced by nutrient availability (Korell et al., 2020;Matthies, 2017;Matthies & Egli, 1999;Těšitel, Těšitelová, et al., 2015), all plant combinations were also grown at two levels of nutrients. I asked the following specific ques- are annual root hemiparasites that attack a wide range of host species (Hautier et al., 2010;Rowntree et al., 2014;Sandner & Matthies, 2018), but can also grow without a host (Matthies & Egli, 1999;, although less vigorously. Seeds of Rhinanthus spp. germinate in autumn at low temperatures, but during the winter, the epicotyl is dormant and only the hypocotyl develops (Hartl, 1974;Westbury, 2004). The seedlings emerge above ground in March to April and grow rapidly, and the parasites start flowering in May. The flowers are pollinated by bumblebees (Kwak, 1979), but may also self-pollinate (Sandner & Matthies, 2017).
Rhinanthus minor is widespread throughout Europe and has also been introduced to North America (Westbury, 2004), where it recently has become invasive (Smith & Cox, 2014), while R. alectorolophus has a more restricted distribution in Central Europe (Hartl, 1974). Both parasite species are mainly plants of grasslands.
While R. minor is a plant of nutrient-poor grasslands, R. alectorolophus grows typically in more nutrient-rich habitats than R. minor (Ellenberg et al., 1992) and was formerly also a weed of cereal crops in Europe (Hartl, 1974).

| The experiment
The hemiparasites R. minor and R. alectorolophus were grown with 25 host species at two nutrient levels ( Table 1). All the host species occur with the Rhinanthus species in their habitats. Seeds of both the parasites and the hosts were obtained from a commercial supplier (Appels Wilde Samen, Darmstadt, Germany). The host species were selected to include species from the functional groups grasses, legumes, and nonleguminous forbs and from nutrient-poor and moderately nutrient-rich habitats. Each species was assigned to one of three groups according to its Ellenberg indicator value for nutrients (Ellenberg et al., 1992). Species with N-values of 2-3 were classified as species of low-nutrient habitats, those with N-values of 4-6 as species of moderately nutrient habitats, and those with N-values of 7-8 as species of high-nutrient habitats. Species for which no Nvalues were available were classified as indifferent.
Seeds of R. minor and R. alectorolophus were set up for germination on moist filter paper in Petri dishes in mid-February and kept at 5°C to break dormancy. At the beginning of May, 10-15 seeds of the host plants were sown into pots of 9 × 9 × 9.5 cm filled with nutrient-poor commercial soil (TKS Instant, Floragard, Oldenburg, Germany), except for seeds of Achillea millefolium, Poa annua, and Lolium perenne, which were known to germinate faster and were sown one week later. The pots were kept in an unheated glasshouse.
After three weeks, the number of seedlings was reduced to three per pot. At the end of May, one seedling of R. minor or R. alectorolophus was transplanted into the center of a number of pots with each host species. After initial transplanting mortality, c. 10 replicates (means: R. minor 9.9, R. alectorolophus: 10.4; range 9-13) per combination of each host and parasite species remained. In addition, ten pots with each host species were left as no-parasite controls. These pots were then placed on saucers in flower beds in the Botanical Garden of the University of Marburg. In a further ten pots with each host species, the biomass of the hosts was harvested above ground, dried for 48 hr at 80°C, and weighed to obtain a measure of initial host size.
During the first two weeks, the plants were protected against the sun with shading cloth that reduced the light intensity by 45%.
Two weeks and four weeks after the planting of the parasites, half of the pots received 40 ml of a 0.3% solution of Wuxal Super (Aglukon, Düsseldorf; N-P-K: 8%-8%-6%) fertilizer, and six weeks after planting, they received another 40 ml of a 0.4% solution of the fertilizer (high-nutrient level) to ensure differences in host growth between the two treatments, while the other pots received only water (lownutrient level).
After four weeks of growth, the length of the longest leaf of each parasite was measured as an nondestructive estimate of size.
Once the parasites started to flower, the date when the first flower opened was recorded for each plant. In the 10th week after planting when the parasites were fruiting, the length of the longest leaf, the height, and total inflorescence length of each parasite were measured. Parasites and hosts were then separately harvested above ground, dried for 48 hr at 80°C, and weighed. Mortality of the parasites was calculated as the proportion of parasites that died after they had survived the first ten days after transplanting.
Fifty seeds of each host species were weighed to obtain its mean seed mass, because host seed mass might influence their early growth rate.

| Statistical analyses
The influence of host species and nutrient level on the mortality of the parasites was analyzed with chi-square tests. The effects of the two treatments on the biomass and other traits of the two parasites were studied by two-factor analyses of variance. Because these traits are not independent, p-values were adjusted for the false discovery rate (Benjamini & Hochberg, 1995). To investigate whether the quality of a host species for the two species of Rhinanthus was correlated, the relationship between the biomass of the two species with the same hosts was studied by linear regression.
Because many of the hosts used in the current experiment had been used in a previous study of the host relations of the related TA B L E 1 Species used as host plants in the experiment. The indicator value for nutrients (N-value, Ellenberg et al., 1992) indicates the realized ecological niche of a species with respect to nutrient level in Central Europe. Species with N-values of 2-3 were classified as species of low-nutrient habitats, those with N-values of 4-6 as species of medium-nutrient habitats, and those with N-values of 7-8 as species of high-nutrient habitats. Species for which no N-values were available, because their behavior is indifferent to nutrients, were classified as indifferent  (Sandner & Matthies, 2018) and one with the parasite Odontites vulgaris (Geppert, 2012).
To analyze the effects of various host traits on the biomass of the parasites, linear mixed models were constructed separately for the two parasite species with host species identity as a random factor and the following fixed factors: nutrient level, nutrient status of the typical habitat of the hosts, mean seed mass of the hosts, mean biomass of the hosts at the start of the experiment, mean final biomass of the hosts grown with or without a parasite, and mean RGR of the hosts. The metric explanatory variables were standardized for the analyses. In a second step, all possible models including the explanatory variables were calculated and ranked by their AICc to obtain the best models. Differences between the performance of the parasites growing with species from different functional groups were then analyzed using Tukey-adjusted p-values.
To analyze the effect of the size of the host plants growing in the same pot on the biomass of the individual parasites, the mean biomass of R. minor and R. alectorolophus was related in general linear models to nutrient level, host species identity, their interaction, and host mass. The mean biomass of the parasites was also related to the length of their longest leaf after four weeks of growth to assess the influence of early differences in size on final biomass.
The effects of host species, parasite species, nutrient level, and their interactions on host biomass and total aboveground productivity (host + parasite) per pot were studied by three-way analyses of variance. To investigate whether damage to a host and benefit to a parasite were correlated, log-response ratios were calculated for the effect of the parasites on the individual host species as log (mean biomass of a host with a parasite/mean biomass without a parasite) and related to the mean biomass of the parasites achieved with the individual host species.
All analyses were carried out with R 4.0.3 (R Core Team, 2020).
Analyses of variance and general linear models were carried out with the package lm. p-Values using type III sums of squares were obtained with the ANOVA function of the car package (Fox & Weisberg, 2019). p-Values adjusted for the false discovery rate were obtained with the p.adjust command. Linear mixed models were carried out with the function lmer of package lme4 (Bates et al., 2015).
All possible linear models and their AICc values using a set of explanatory variables were calculated with the dredge function of the MuMIn package (Barton, 2020). Mean values and Tukey-adjusted pvalues were obtained with the emmeans package (Lenth, 2020). Data for biomasses, height, inflorescence length, and seed mass were logtransformed prior to analysis to obtain normally distributed residuals and homoscedasticity.

| Influence of the host species on parasite traits
There was no evidence that the mortality of the two hemiparasites (R. minor: 41%, R. alectorolophus: 19%) was influenced by the host species (R. minor: χ 2 = 16.5, df = 24, p = 0.87; R. alectorolophus: The biomass of both hemiparasites was strongly influenced by the identity of the host species, and that of R. alectorolophus also by nutrient level, but the interaction between the two factors was far from significant (  (Benjamini & Hochberg, 1995) were poorer hosts than expected. The quality of a certain species as a host for R. alectorolophus in the present experiment was also strongly correlated with that of the same species in another experiment (Sandner & Matthies, 2018) with R. alectorolophus and several of the same host species (r = 0.72, p = 0.012; Figure 3b). In contrast, the quality of a species as a host for R. alectorolophus and that for the related hemiparasite Odontites vulgaris in a previous experiment (Geppert, 2012) with some of the host species was less strongly correlated (r = 0.49, p = 0.017; Figure 3c), and the correlation with the performance of Melampyrum arvense in another experiment (Matthies, 2017) was even weaker (r = 0.21, p = 0.34; Figure 3d). This was mainly due to a number of grasses that were good hosts for Rhinanthus, but not for Melampyrum (Trisetum, Dactylis, Bromus, Cynosurus, Koeleria), and a number of forbs that were good hosts for Melampyrum, but rather poor hosts for Rhinanthus (Capsella, Urtica, Taraxacum, Achillea).
The biomass of R. alectorolophus with all host species was much higher than that of R. minor, on average by 370%. Higher nutrient levels increased the biomass of R. alectorolophus by 87% while the effect on the biomass of R. minor (+47%) was not significant (

| Effects of host traits on parasite biomass
In AICc. The ΔAICc between these best models and the best models including further explanatory variables was 3.0 (R. minor) and 5.2 (R. alectorolophus).
In the best models, host quality as measured by the mean biomass of the parasites differed strongly among the three functional groups grasses, legumes, and nonleguminous forbs for both R. minor (χ 2 = 32.6, p < 0.001) and R. alectorolophus (χ 2 = 30.0, p < 0.001).
The mean biomass of both R. minor and R. alectorolophus grown with a legume (+347% and +463%; both p adj < 0.001) or with a grass (+209% and +202%; p adj < 0.001 and p adj = 0.002) was much higher than when grown with a nonleguminous forb (Figure 4)

| Influence of the parasites on the growth of the host plants and total productivity
Biomass varied among the 25 host species and was higher at highnutrient levels (  The effect of the parasites on host growth did not depend on nutrient level, but there was strong evidence for R. alectorolophus and much weaker evidence for R. minor that the effect of the two parasites on the hosts varied among species (Table 3, Figure 5). The mean reduction in the biomass of a host species by the two parasites as measured by the log-response ratio of their biomass to parasite presence was not related to the mean biomass of the parasites (R. minor: r = 0.13, p = 0.53; R. alectorolophus: r = 0.11, p = 0.60), that is, damage to a host was not related to the benefit it provided for the parasites (Figure 5a,b). Some species such as Papaver, Taraxacum, A few species were more strongly damaged by R. minor (e.g., Myosotis, Medicago sativa, and M. lupulina), but most hosts were more strongly negatively affected by R. alectorolophus than by R. minor ( Figure 5c). However, hosts that were strongly damaged by one of the parasites tended also to be strongly damaged by the other one, as shown by the positive correlation between the log-response ratios of the effects of the two parasite species on the same hosts (r = 0.38, p = 0.061). The three functional groups grasses, forbs, and legumes did not vary in the mean degree of biomass reduction due to R. minor (F 2,22 = 1.66, p = 0.21) and R. alectorolophus (F 2,22 = 1.44, Total aboveground productivity per pot (host + parasite biomass) varied depending on the host species (Table 3) and was on average 26% higher at high-nutrient levels. The effect of the presence of a hemiparasite on productivity also differed between the two parasite species and the hosts, while it was hardly influenced by nutrient level. Separate analyses for the two parasites showed that R. minor tended to slightly reduce overall productivity per pot (−5%, F 1,296 = 2.86, p = 0.092). This effect did not differ among host species (F 24,296 = 0.88, p = 0.63). In contrast, the overall effect of the presence of R. alectorolophus on total aboveground productivity was positive (+9%, F 1,360 = 8.71, p = 0.003), but differed strongly depending on the host species (F 24,360 = 2.77, p < 0.001; Figure 6). Although the parasite R. alectorolophus reduced the biomass of nearly all hosts, it nevertheless increased total aboveground biomass per pot with many species, in particular Poa, Koeleria, Lolium, Trifolium repens, and T. pratense, because its own biomass production more than compensated for the loss of host mass by parasitism (upper left quadrant in Figure 6). However, with other host species such as Arrhenatherum or Chrysanthemum, the parasite reduced both host mass and total productivity per pot (lower left quadrant in Figure 6).

| Influences on the performance of the two hemiparasite species
The performance of the two hemiparasites R. minor and R. alectorolophus grown with the same host species was closely correlated indicating that at least for two congeneric species of root hemiparasites grown under the same conditions, the relative quality of individual species as hosts is very similar. Moreover, the performance of R. alectorolophus in the present experiment was also significantly correlated with that of R. alectorolophus in a previous experiment involving many of the same hosts (Sandner & Matthies, 2018) and

F I G U R E 4 Biomass of the hemiparasites Rhinanthus minor and
with that of the related hemiparasite Odontites vulgaris in another experiment (Geppert, 2012), although to a lesser degree. Significant correlations between the performance of the hemiparasites were found although the three experiments differed in conditions such as the number of host individuals per pot, soil type, and nutrient levels, which are known to influence hemiparasite-host interactions. In contrast, the performance of Rhinanthus was poorly correlated with that of the hemiparasite Melampyrum arvense grown with many of the same species (Matthies, 2017 in previous studies (Barham, 2010;Cameron & Seel, 2007;Rowntree et al., 2014), which has been attributed to its defense reaction to parasite attack (Cameron & Seel, 2007

F I G U R E 6
The relationship between the reduction in total aboveground biomass (parasite and host) and the reduction in host biomass for the hemiparasite Rhinanthus alectorolophus. Shown are log-response ratios = log (mean biomass with a parasite/mean biomass without the parasite). Negative log-response ratios indicate that host biomass or productivity is reduced by the hemiparasite. For abbreviations of host names, see Table 1 observed in other studies of the performance of Rhinanthus spp.
with different hosts (Hautier et al., 2010;De Hullu, 1984;Rowntree et al., 2014; and higher than the range observed for other hemiparasites (Calladine et al., 2000;Guo & Luo, 2010;Marvier, 1996;Radomiljac, 1999), but lower than the range found for Melampyrum arvense when grown with 25 host species (171-fold at low nutrients; Matthies, 2017). The large variation in the performance of Rhinanthus in the present experiment can be related to the large number of host species used which increased the probability that both very poor and very good hosts were among the species studied. Indeed, the variation in the performance of Rhinanthus with different hosts species in previous studies increased with the number and diversity of hosts studied: The biomass of R. alectorolophus grown with 9 hosts (only grasses) varied 2.4-fold (Hautier et al., 2010), that of R. minor grown with 9 grasses and forbs 7-fold (Rowntree et al., 2014), that of R. minor grown with 11 species 13fold , that of R. alectorolophus with 13 species 11fold (Sandner & Matthies, 2018), and that of R. angustifolius grown with 18 species 20-fold (De Hullu, 1984). However, differences in growth conditions and the length of the growth period may also have contributed to the differences among studies.
The quality of a species as a potential host for a root hemiparasite may depend on many aspects of their interactions. Parasites may not attach to the roots of a potential host because it does not stimulate haustoria formation, the haustoria may have problems to penetrate the roots due to the structure or thickness of the roots, and host roots may defend themselves by blocking haustoria (Cameron et al., 2006;Govier, 1966;Yeo, 1964). Moreover, the quantity and quality of compounds provided by a host may vary between species, as grasses have been found to provide mainly carbon while legumes provided nitrogenous compounds (Govier et al., 1967). The ability of host shoots to capture more of the resources taken up by the host roots than the parasite may also vary among species, as well as the impact of the host on hemiparasite growth by shading (Matthies, 1995a). The variation in growth of the two species of Rhinanthus with the different host species could not be explained by several characteristics of the host species that were found in other experiments to have an influence, like their size when grown with or without a parasite, their growth rate or realized niche with respect to nutrients (Hautier et al., 2010;Marvier, 1996;Matthies, 2017). Thus, large hosts did not provide more solutes to the parasites than small ones and fast-growing species were not better hosts than slowgrowing ones.
In contrast, there were significant differences among functional groups in their quality as host plants. Of the 25 species studied, the five legumes and nine grasses were on average much better hosts for both R. minor and R. alectorolophus than the eleven nonleguminous forbs. The often observed large benefits of legume hosts for hemiparasites (Lu et al., 2014;Radomiljac, 1999;Rowntree et al., 2014;Tennakoon & Pate, 1996;Yeo, 1964) could be attributed to their symbiosis with nitrogen-fixing bacteria and consequent high supply of nitrogen (Govier et al., 1967, Jiang et al., 2008. Similarly, a strong preference for potentially N-fixing species as hosts (legumes and nonlegumes) was found for the hemiparasite Santalum acuminatum (Tennakoon, Pate & Arthur, 1997).
The suitability of grasses has been attributed to the weak defense of their roots against parasite attack (Cameron et al., 2006;Rümer et al., 2007). However, these general effects masked considerable variation within the functional groups. The performance of the Rhinanthus with forbs as hosts was poor, but this was also the case with several of the grasses such as Anthoxanthum, Bromus, Arrhenatherum, and Cynosurus. In a comparison of nine host species, Rowntree et al. (2014) also found forbs to be the least beneficial hosts for R. minor. In contrast, in a study of the growth of R. alectorolophus with 13 host species (Sandner & Matthies, 2018), grasses were the best and legumes the worst hosts. However, this was due to the presence of two legumes (Anthyllis and Onobrychis) in the experiment that were very poor hosts. Thus, while most legumes generally appear to be good and forbs rather poor hosts for Rhinanthus as concluded by Cameron and Phoenix (2013), individual species may deviate from this general pattern. Why some legumes were poor hosts in studies of hemiparasite-host interactions is usually not clear, but Govier (1966) found that Trifolium incarnatum blocked the haustorium of the hemiparasite Odontites verna by producing a layer of a substance (probably tannin) between the appressorial cells of the haustorium and the stele of the clover.
The identity of the host species influenced also traits of the parasites other than biomass such as their leaf length, height, inflorescence length, and time until flowering, but this was to a large degree an effect of the effect of host identity on parasite size. Parasites attached to good hosts grew faster and started to flower earlier.
Pollinations early in the season may thus occur mainly between parasite individuals that have been successful in parasitizing certain host species that are most beneficial. As the ability to successfully exploit individual host species has a genetic component (Ahonen et al., 2006;Rowntree, 2014;Sandner & Matthies, 2017), this would result in assortative mating and might facilitate the evolution of genotypes adapted to specific hosts. However, there is yet little evidence for the evolution of specialization on hosts in Rhinanthus (Ahonen et al., 2006;Mutikainen et al., 2000).
Higher nutrient levels did not affect the survival and increased the growth of both species of hemiparasite, although in the case of R. minor, this effect was not significant. Nutrient levels in the current study were thus not sufficient to change the balance in the competition for light between the hemiparasites and their hosts in favor of the host plants which would have resulted in increased mortality of young hemiparasites (Matthies, 1995;Matthies & Egli, 1999;Mudrák and Lepš, 2010;Těšitel, Těšitelová, et al., 2015). Because the host plants are for hemiparasites simultaneously beneficial sources of water, nutrients, and carbon, but also competitors for light, hemiparasites are restricted to habitats of low-nutrient availability (Matthies, 1995;Těšitel, Fibich, et al., 2015). However, in experimental studies using pots negative effects of high-nutrient levels on hemiparasites are not always observed (Borowicz & Armstrong, 2012;Korell et al., 2020;Matthies & Egli, 1999) and will depend on maximum nutrient levels, host density, and host age.

| Effects of the two hemiparasite species on the hosts
The negative effects of root hemiparasites on the growth of their hosts are often very strong (Hautier et al., 2010;Korell et al., 2020;Matthies, 2017;Matthies & Egli, 1999; but see Tennakoon, Pate & Fineran, 1997). In a meta-analysis, Ameloot et al. (2005) concluded that Rhinanthus spp. on average reduced host biomass by 60% in pot and 40% in field experiments. In later experiments, R. alectorolophus reduced the mass of its host species by between 9% and 37% (Sandner & Matthies, 2018), and by 56% (Korell et al., 2020), and R. minor by 26% (Bardgett et al., 2006). In the current study, host damage was far less severe. The mean reduction of host biomass was 12% by R. minor and 19% by R. alectorolophus, and even the biomass of the most strongly affected hosts was only reduced by 36% (R. minor) and 45% (R. alectorolophus). The lower impact of the parasites on their hosts could be due to the fact that three host individuals were planted with each parasite per pot, thus potentially reducing the effect of the parasite (but see Korell et al., 2020). Moreover, the plants were well watered and the negative effects of hemiparasites on their hosts may be strongest if either water or nutrients are strongly limiting plant growth, as water and nutrients are the two key resources that root hemiparasites extract (Těšitel, Těšitelová, et al., 2015).
The negative impact of the parasites on the growth of the hosts differed strongly among host species, but the effect of the two Rhinanthus species was correlated, although there was a lot of variation, indicating that the two parasite species made similar relative demands on the hosts. The fact that a host species suffers no or little damage by a parasite can be due to constitutive resistance of the root system of a host to parasitic attack or to a successful defense reaction (Atsatt, 1983;Cameron & Seel, 2007). This is the likely explanation for the low damage in species that provided little benefit to the parasites such as Hieracium and Papaver. However, several legumes (Medicago lupulina, M. sativa, Trifolium pratense, T. repens, Lotus) and the grass Dactylis supported large parasites of one or both Rhinanthus species and were thus good hosts, but were hardly suppressed by the parasites. These species were thus tolerant of parasitism. Tolerance against parasitism in Lotus and Trifolium pratense has also been observed in another experiment with R. alectorolophus (Sandner & Matthies, 2018), and tolerance against the related hemiparasite Melampyrum arvense was found for several of the same legume species as in the current experiment (Lotus, Trifolium pratense, T. repens; Matthies, 2017). All these species grew vigorously and produced a lot of biomass, and vigorous growth is also a typical constitutive trait of plants that are tolerant of the attack of herbivorous insects (Fornoni, 2011). Tolerance of the legumes may have been facilitated by their mutualistic symbiosis with nitrogenfixing Rhizobia. The fact that several host species were tolerant of parasitism could partly explain the low correlation between the benefit a parasite derived from a certain host species and the damage in terms of reduced biomass it caused the host. This weak relationship indicates that the damage to the host cannot be explained simply by the amount of resources extracted by the hemiparasites. In contrast, in Melampyrum arvense the benefit of a host species for the parasite and the damage to this host have been found to be strongly correlated (Matthies, 2017).
It has been suggested that the negative impact of hemiparasites on their hosts will be particularly strong at low levels of nutrient availability (Matthies, 1995a;Těšitel, Těšitelová, et al., 2015), because under high-nutrient conditions the loss of nutrients to the parasite will be less detrimental to the host plants. Some studies have found support for this notion (Liu et al., 2017;Matthies & Egli, 1999;Těšitel, Těšitelová, et al., 2015), but others found no influence of nutrient level on host damage by hemiparasites (Bardgett et al., 2006;Korell et al., 2020;Mudrák and Lepš, 2010) or only very small effects (Matthies, 2017). Similarly, no increased damage to the hosts at lownutrient levels was observed in the present study. These conflicting results indicate that variation in other factors strongly influences the effect of nutrients on hemiparasite-host interactions.

| Effects on total productivity
Because hemiparasites have a lower resource use efficiency than their hosts and affect host photosynthesis, negative effects of the presence of hemiparasites on total productivity (host and parasite combined) can be expected (Hautier et al., 2010;Matthies, 1995).
To obtain nutrients from the roots of their host plants, hemiparasites have very high rates of transpiration, even in the dark when autotrophic plants close their stomata (Lechowski, 1996;Press et al., 1988), and while parasites may accumulate very high concentrations of nutrients in their tissues (Pate et al., 1990), their own rates of photosynthesis are similar to or lower than those of their hosts (Lechowski, 1996;Press et al., 1993). Hautier et al. (2010) even predicted based on a model that the presence of hemiparasites will always reduce total productivity. However, the results of empirical studies on the effect of hemiparasites on total productivity are inconsistent. Some studies have found a reduction of overall productivity due to the presence of a hemiparasite (Hautier et al., 2010;Korell et al., 2020;Matthies, 1995aMatthies, , 1995bMatthies, 1996;Matthies & Egli, 1999;Mudrák & Lepš, 2010), while others have found an increase of productivity with at least some hosts (Matthies, 2017;Sandner & Matthies, 2018) or host combinations (Joshi et al., 2000;Sandner & Matthies, 2018).
In the present experiment, the two species of Rhinanthus had different effects on productivity. Total aboveground productivity was on average reduced by R. minor, but actually increased by R. alectorolophus, indicating that in many parasite-host combinations the loss in host biomass due to parasitism was more than compensated by the carbon gain through the photosynthesis of R. alectorolophus.
However, because root hemiparasites invest very little biomass into their own roots and instead rely on the resources taken up by the roots of their hosts (Matthies, 1995(Matthies, , 2017, the negative effect of the parasites on total productivity (above ground and below ground) could have been stronger.

| CON CLUS IONS
The performance of the two hemiparasites R. minor and R. alectorolophus grown with the same host species was strongly correlated, and also correlated with that of R. alectorolophus and Odontites in other experiments, in spite of differing conditions. Moreover, the two Rhinanthus species had related effects on the host species. This shows that the interactions between closely related root hemiparasites and individual host species are similar and that certain root traits and defense mechanisms of potential hosts may be effective against several hemiparasites. However, the weak correlation between the performance of Rhinanthus and Melampyrum with the same hosts shows that there are also specific interactions between pairs of hemiparasite-host species. The differences in the response of Rhinanthus and Odontites versus that of Melampyrum could be related to their phylogenetic relationships, as the genus Odontites is more closely related to Rhinanthus than is Melampyrum, which forms a sister group to other Orobanchaceae (Těšitel, Říha, et al., 2010).
The results of this study confirmed that legumes are in general, although not universally, very good hosts for hemiparasites and that some of them are tolerant of parasitism (Matthies, 2017;Sandner & Matthies, 2018). Both phenomena can be related to their symbiosis with nitrogen-fixing Rhizobia, because the high nitrogen content of legumes benefits the parasites, while legumes will also be more capable to compensate for the loss of nitrogen to the parasites than plants from other functional groups (Matthies, 2017).
Both R. minor and R. alectorolophus have been used as model species for the study of host-hemiparasite relationships. However, the present study revealed important differences between the two parasites. The biomass of R. alectorolophus was much higher than that of R. minor with every host species and R. alectorolophus also reacted more strongly with increased growth to nutrient addition than did R. minor, indicating that R. alectorolophus is the more competitive species in fertile habitats. The results are in line with the different habitats of the two parasites. While R. minor is a typical species of nutrient-poor grasslands, R. alectorolophus grows in more mesotrophic grasslands and also formerly occurred as an agricultural weed (Ellenberg et al., 1992;Hartl, 1974). The fact that R. alectorolophus in contrast to R. minor increased overall productivity and caused less damage to the hosts in relation to its own biomass shows that R. alectorolophus has a greater capacity to use nutrients obtained from the host to increase its own photosynthesis, makes less demands on its hosts, and is thus less parasitic than R. minor. In line with this, carbon taken up from the host accounted for 50% of total carbon in R. minor (Těšitel, Plavcová, et al., 2010) but only for 10%-40% in R. alectorolophus (Těšitel, Těšitelová, et al., 2015), depending on host and growth conditions.

ACK N OWLED G EM ENTS
I thank Christine Krebs for technical assistance and Tobias Sandner for comments on an earlier version of the manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data from this study are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.n5tb2 rbw8.