Competition with wind-pollinated plant species alters floral traits of insect-pollinated plant species

Plant traits related to attractiveness to pollinators (e.g. flowers and nectar) can be sensitive to abiotic or biotic conditions. Soil nutrient availability, as well as interactions among insect-pollinated plants species, can induce changes in flower and nectar production. However, further investigations are needed to determine the impact of interactions between insect-pollinated species and abiotically pollinated species on such floral traits, especially floral rewards. We carried out a pot experiment in which three insect-pollinated plant species were grown in binary mixtures with four wind-pollinated plant species, differing in their competitive ability. Along the flowering period, we measured floral traits of the insect-pollinated species involved in attractiveness to pollinators (i.e. floral display size, flower size, daily and total 1) flower production, 2) nectar volume, 3) amount of sucrose allocated to nectar). Final plant biomass was measured to quantify competitive interactions. For two out of three insect-pollinated species, we found that the presence of a wind-pollinated species can negatively impact floral traits involved in attractiveness to pollinators. This effect was stronger with wind-pollinated species that induced stronger competitive interactions. These results stress the importance of studying the whole plant community (and not just the insect-pollinated plant community) when working on plant-pollinator interactions.

of plant communities can have strong effects on individual plant traits through competitive interactions for resources.
Variations in plant reproductive traits are especially important in animal-pollinated species because they condition plant-pollinator interactions 17,18 . Indeed, in insect-pollinated plants, pollinators are attracted to flowers and associated rewards. On the one hand, flowers, varying in their number, size, color or smell, offer an advertising display that induces visits of various pollinators 17,19 . On the other hand, floral rewards (e.g. nectar, pollen) tend to favor the repetition of visits as they are major components of pollinators' diet by supplying proteins, sugars and amino-acids 17 . The quantity and quality of such floral traits involved in pollinator attraction can have strong impacts on pollinator behavior. Indeed, several studies have shown that plant species exhibiting a greater floral display size, (i.e. the total number of opened flowers at a time) or producing numerous, large flowers and/or greater rewards (in quantity or quality) are more visited by pollinators than other present plant species [20][21][22][23][24][25] . Various experiments showed that a greater pollinator attractiveness, subsequently to increased floral traits through resource addition, can enhance pollinator visitation [6][7][8]26 that may lead to better reproductive success 6,8 . However, to date, most studies linking changes in soil resources to floral traits and pollinator response considered the impact of nutrient addition while interactions between plants, especially belowground competition, could also be of great importance. In an experiment looking at different floral traits involved in attractiveness, Baude et al. 11 set up binary mixtures of insect-pollinated plant species and found that floral traits of one focal species depended on the other species present in the mixtures. More precisely, the total nectar sugar content of a focal species decreased when growing in presence of a stronger competitor. Therefore, interactions between insect-pollinated species could influence floral traits involved in pollinator attraction.
Natural plant communities always comprise species with a variety of pollination modes such as animal-pollinated and abiotically pollinated plants, although the latter are almost never taken into account in studies on plant-pollinator networks. However, the flower production of a particular insect-pollinated species can be negatively impacted by competitive interactions induced by a wind-pollinated plant competitor 15,27 . To our knowledge, the consequences of interactions between insect-and wind-pollinated species on floral traits involved in attractiveness to pollinators still need investigations, especially with more focus on floral rewards.
The objectives of our study were to assess how allocation to several floral traits involved in attractiveness to pollinators (e.g. flower production, flower size and floral rewards) could be affected by the presence of different wind-pollinated species. Especially we wanted to investigate the effect of different intensities of competitive interactions. To do so, we set up a pot experiment in which we grew three insect-pollinated plant species (Echium plantagineum, Lamium purpureum and Lotus corniculatus) in binary mixtures with four wind-pollinated species (Agrostis capillaris, Chenopodium album, Holcus lanatus and Plantago lanceolata) so that insect-pollinated species were submitted to a panel of belowground interactions for abiotic resources.
Our hypothesis were that (1) the presence of wind-pollinated competitors should have negative impacts on floral traits of insect-pollinated species and (2) the magnitude of this effect should differ according to different competition intensities induced by the presence of wind-pollinated species.

Results
Intensity of competitive interactions. Mean log response ratios (ln RR 28 ), estimators of competition intensity, are summed up in Table 1. As indicated by ln RR values, the three insect-pollinated focals were submitted to various intensities of competition. E. plantagineum and L. purpureum followed the same pattern of response. For both species and whatever the biomass measurement (aboveground, belowground or total biomass), ln RR values (as well as focals' biomass) were significantly higher in presence of C. album, in opposition to mixtures with H. lanatus for which ln RR had the lowest values (P ≤ 0.024 for both species, Table 1, Supplementary Fig. S1 for biomass). Mixtures with C. album even seemed to provide better growth conditions than monocultures of the two focal species with positive values of ln RR (Table 1). For intermediate levels of competition intensity, ln RR values suggest a global pattern where intensity of competition is stronger in mixtures with A. capillaris than with C. album but weaker than with P. lanceolata, the second strongest competitor for L. purpureum and E. plantagineum (except for ln RR values calculated with belowground biomass of L. purpureum, Table 1). According to these results, L. purpureum and E. plantagineum experienced the following panel of growing intensity of competition: The response of L. corniculatus didn't follow the same pattern. Monocultures showed the highest ln RR values among all treatments (Table 1). In consequence, all wind-pollinated species induced negative competitive interactions for this focal species ranging from weak effects in mixture with C. album to strong effects in mixture with P. lanceolata or A. capillaris (Table 1).
As we are mainly interested in the effect of competition intensity induced by wind-pollinated species rather than on competitor identity, the log response ratio (ln RR) has been used as an explanatory variable in the following results. However, because ln RR values were obtained from biomass measurements at the end of the experiment, we used ln RR (calculated from total biomass, see in Methods) as an explanatory variable only for total floral traits (i.e. summed at the end of the experiment, see Flower traits, Nectar traits and Data analysis sections in Methods). We kept the competitor identity as an explanatory variable for daily floral traits see Flower traits, Nectar traits and Data analysis sections in Methods). Therefore, in the following results, 'competition intensity' will equally refer to 1) ln RR values, 2) the presence of a particular competitor in the mixture with the focal species. An increase in competition intensity can thus mean a decrease in ln RR values or the presence of stronger competitors. For the sake of clarity, the results for daily floral traits will be ordered along the above panel for the three focal species, even if the pattern of response was different for L. corniculatus.

Flower traits.
At the end of the experiment, a total of 807, 2053 and 1075 flowers were sampled for E. plantagineum, L. purpureum, and L. corniculatus, respectively. For E. plantagineum and L. purpureum there was a significant effect of competition leading to a decrease of floral display size, daily flower production (not shown) and total flower production (total number of flower produced at the end of the flowering period, see Methods) when competition intensity increased. Indeed, floral display size and daily flower production tended to be greater in presence of C. album (even greater than in monocultures; although not significant) while the presence of stronger competitors, such as H. lanatus, induced a strong decrease in both traits ( Fig. 1 for floral display size, the daily flower production followed the same pattern). Likewise, for both species, the total flower production decrease as ln RR values decrease, suggesting lower total flower production in condition of stronger competitive interactions (P < 0.001 Supplementary information Fig. S2). For L. corniculatus, there was also a significant effect of the competition treatment. Especially, floral display size and daily flower production were greater in presence of C. album ( Fig. 1 for floral display size, the daily flower production followed the same pattern). However, both traits tended to be lower in monocultures ( Fig. 1 for floral display size, the daily flower production followed the same pattern). On the other hand, the total flower production decreased according to ln RR values (P < 0.001, Supplementary information Fig. S2). For all three species, there were significant effects of the date (for the floral display size only, all P < 0.001) and of the interaction terms for both floral display size and daily flower production (all P < 0.001).
For all three focal species, flower size was affected by the competition treatment (E. plantagineum: As the daily flower production was affected by the competition treatment, the daily sucrose index of nectar (taking into account all produced flowers per day and not only sampled flowers, see Methods) decreased in presence of strongest competitors for both E. plantagineum and L. purpureum (F 4,194 = 13.36, P < 0.001and F 4,731 = 5.52, P < 0.001 respectively). Therefore, the daily allocation of sucrose to nectar was lower when competition intensified. The same pattern was clearly observed for the total allocation of sucrose to nectar. Indeed, the total sucrose index at the end of the flowering period tended to decrease with ln RR values (E. plantagineum F 1,47 = 57.04, P < 0.001, L. purpureum: F 1,48 = 83.81, P < 0.001; Fig. 2, see Methods). We found the same pattern for daily (E. plantagineum F 4,199 = 9.38, P < 0.001; L. purpureum F 4,800 = 8.05, P < 0.001) and total volume index (F 1,46 = 30,71, P < 0.001, F 1,48 = 102,74 P < 0.001 respectively). For L. corniculatus, the daily sucrose index and the daily volume index were significantly affected by the competition treatment (F 4,255 = 3.48, P = 0.009 and F 4,254 = 3.47, P = 0.009 respectively) with lower values in monocultures. However both index of nectar were not affected by increasing competition intensity (both P > 0.05, see Fig. 2 for total sucrose index). There were significant effects of the date for the daily concentration of nectar (P ≤ 0.02 for E. plantagineum and L. purpureum) and the daily nectar  volume (L. purpureum only P < 0.001). There was a significant effect of the interaction term (competition:date) for the daily nectar volume (for E. plantagineum and L. corniculatus, all P ≤ 0.01). For both daily index, there were significant effects of the date (for L. purpureum only, P < 0.001) and of the interaction term (competition:date) (for E. plantagineum and L. corniculatus, all P ≤ 0.002).

Discussion
Literature data have mostly focused on the relations between plant attractiveness to pollinators and abiotic conditions and suggest that response of attractiveness traits is complex and species-specific as positive, neutral and negative effects have been reported for the effects of water or nutrient addition on flower productionp 3,7,11,12 . Because competitive interactions between plants can translate into changes in resources availability between competitors 13,29 , we studied their impact on floral traits of three insect-pollinated species. Moreover, we focused on wind-pollinated species as competitors in order to see how species that do not interact with pollinators for their reproduction could alter attractiveness traits of animal-pollinated focal species.
As expected, competition with wind-pollinated species affected plant biomasses and floral traits for all three investigated insect-pollinated species. Especially, we found that the stronger the competitive interaction was (i.e., high ln RR values), the stronger the impact on floral traits.
Differences in biomass allocation patterns, especially belowground, can suggest different competitive abilities in plants 29 . Indices of competition, such as the log response ratio (ln RR) are frequently used as they are good tools for summarizing and interpreting competitive interactions between plant species 28 . Here, the ln RR values obtained for each mixture indicate that the wind-pollinated competitors exposed the insect-pollinated focals to different competition intensities. These ln RR values also suggest that the two annual species of our experiment, L. purpureum and E. plantagineum, faced a similar panel of growing intensity of competition, with C. album being a weaker competitor, A. capillaris an average competitor and P. lanceolata and H. lanatus being stronger competitors. Measures of competitors' biomass grown in mixtures with L. purpureum or E. plantagineum could explain these differences. Indeed, in these mixtures, plants of H. lanatus produced the greatest biomass among the four competitor species. Even though greater biomasses, especially greater root systems, are not always associated with greater competitive abilities 30 , larger root systems can be related to greater soil space occupation 31 and/ or greater resource uptake 29 thus limiting access to resources for neighbouring plants. Likewise larger individual plants can induce stronger effects on target plants than smaller ones 32 . Here, H. lanatus is a strong competitor (so, with high competitive abilities) because its presence (its biomass) probably led to a strong limitation of the biomass production of the two focal species (lower ln RR values) through strong space occupation and/or greater nutrient depletion. In contrast, C. album individuals tended to have a positive effect on the biomass of these two focals species. Facilitative interactions between plants can be observed through modifications of soil components (e.g. moisture, nutrients) 33 or enhancement of seedling establishment. However here, the positive effect of C. album may probably be due to low biomass production rather than facilitation. This may have favoured greater space/nutrient exploitation by the two focal species, and thus higher allocation to biomass than in the other treatments (including focals monocultures). So in this study, we consider that the competitive abilities of plant species are more a consequence of their biomass production (even if biomass production can also result from higher competitive abilities). In the case of L. corniculatus, response patterns were different. Mixture with C. album appart, H. lanatus behaved as an intermediate competitor in spite of its important biomass production (especially root biomass). Moreover, even though the range of ln RR values was narrower for this species (compared to L. purpureum for example), all wind-pollinated species had a small negative effect on L. corniculatus, compared to monocultures, and the strongest competitors were P. lanceolata and A. capillaris. However, biomass measurements indicate that only L. corniculatus belowground biomass was altered by the presence of a competitor and aboveground biomass was unaffected (see Supplementary information Fig. S1). Some characteristics of L. corniculatus could have mediated this different response to competition compared to the two other species. First, L. corniculatus is a legume species and, although we did not quantify them, L. corniculatus roots showed nodules, indicating that nitrogen fixation did occur in our experiment. L. corniculatus could thus have accessed to the atmospheric N pool 34 so that it was only slightly affected by competition compared to the two other species. Furthermore, L. corniculatus has a perennial life cycle that can induce a different timing of response to competition as well as different allocation patterns compared to plants having annual life cycles 3,35 . Initially, the experimental design contained a second perennial plant, Mimulus guttatus DC. Moreover, we should keep in mind that these conclusions rely on final harvests of biomass when some studies suggest regular harvest along experiments to better assess the dynamics of competitive interactions 36 .
Most of the floral traits measured in this experiment were affected by the competition treatment. Higher conditions of competition (lower ln RR values induced by the presence of H. lanatus) had the greatest impact on E. plantagineum and L. purpureum, by reducing flower and reward traits. To date, studies that have looked for links between attractiveness traits and environmental conditions have mostly focused on the effects of abiotic conditions and showed a sensitivity of attractiveness traits to nutrients and water availability 3,6,9,12 or litter and compost additions to soil 11,26 . If modifications of abiotic conditions can alter species attractiveness to pollinators, it is not surprising that biotic interactions such as competition, that mediate abiotic resources availability, have similar effects 15 . Flower and nectar production can be relatively costly for a single plant [37][38][39] so that allocation to reproductive structures might be modified in a context of competition with limited access to nutrients. Here, the lowest ln RR values, calculated from mixtures in presence of H. lanatus, suggest that this species may have reduced the availability of soil resources to E. plantagineum and L. purpureum and thus daily as well as total allocation of plants to floral traits. Overall, this could be responsible for lower resources allocation to reproductive traits. Conversely, C. album, led to higher flower and nectar production than in monocultures. A greater resources availability or space, due to the reduced biomass of the competitor might have led to better growing conditions for the insect-pollinated species, resulting in better resource acquisition (as confirmed by higher ln RR values) and increased allocation to reproductive structures.
In the case of L. corniculatus, while ln RR values suggest stronger competition (albeit limited) in presence of wind-pollinated species compared to monoculture, some floral traits were lowest in monoculture and in mixtures with P. lanceolata (which is the strongest competitor for L. corniculatus based on ln RR values) compared to the other mixtures. This suggests that, in contrast to the two other insect-pollinated species, allocation to reproduction was not related to biomass allocation. In Wurst & Van Beersum 40 , monocultures of L. corniculatus can have higher biomass and produce more flowers than in mixture with H. lanatus, which is not in accordance with the observed pattern here. Considering the cost of N 2 fixation suggested in some studies (in term of C allocation to symbiont 41 ), plants in monocultures in our experiment might have allocated less photosynthetates to floral traits leading to the observed decrease in monocultures. However, the study of floral traits per unit of biomass (total floral traits divided by the final biomass) revealed that L. corniculatus might have a more adaptive response to competition while the two other focal species might have a 'purely' plastic response to resource availability. Indeed, for L. corniculatus, floral traits per unit of biomass tend to be higher when competition intensity increases, showing a possible strategy to better attract pollinator in condition of competition. However, as we only have biomass data at the end of the experiment (and a final biomass can not only be considered as a sum of biomass like for total produced flowers, for instance), we believe that further investigations are needed to conclude on these effects.
Here we focused on flower and nectar production while other attractiveness traits could also be affected by competitive interactions. For example, plant pigments or volatile compounds involved in flower colours 42 and scents, relative amounts of different sugars or amino-acids content in nectar 12,43 and pollen quantity and or quality 9 , are all sensitive to resource variations, and could be affected by competitive interactions. Even though we observed a negative impact of competition on some floral traits involved in attractiveness to pollinators, the response of floral traits can be complex and species-specific 3,11,12 . Moreover, we interpret our results in a context of exploitative competition through soil resources depletion while other competitive mechanisms (e.g. interference through allelochemicals) 13,29 could conjointly influence plant response. As ln RR values did not differ among total or belowground biomass and root competition is often stronger than shoot competition (especially with grass competitors 44 ), our results are mainly interpreted in a context of belowground competition. However further investigations are needed to better assess the overall impact of plant competition (aboveground as well as belowground) on floral traits involved in attractiveness to pollinators.
Variations in attractiveness traits are known to strongly impact pollinator visitation patterns and on a larger scale pollination service. Indeed, greater plant attractiveness can enhance the frequency or number of flower visits: most pollinators are preferentially attracted to plants producing numerous, large flowers and/or greater rewards (in quality or quantity) [20][21][22][23][24][25]45 . Larger floral display size can also influence the abundance of visiting pollinators 46 . Likewise, the pattern of pollinator visits per plant can be correlated to the total nectar production per plant 47 . However, many flowering plants are pollen limited, therefore an increase in pollination intensity (e.g. a greater pollen deposit on stigmas) can enhance plant fecundity (i.e. greater fruit and/or seed set) 48 . As a consequence, our results suggest that a decrease in floral traits involved in pollinator attractiveness due to plant competition could have negative impacts on pollinator visits, reducing plant reproductive success. However further experiments are needed to test such hypothesis. Nevertheless, this study emphasises the importance of 1) taking into account species other than insect-pollinated ones in plant-pollinator network studies, and 2) linking above ground and below ground interactions to better understand plant-pollinator networks. This is in concordance with some research initiated on the impact of soil micro-organisms on pollinator visits through variations of floral traits 49,50 . Given our results, future research is needed on plant-soil or plant-plant interactions that may lead to modifications of floral traits involved in attractiveness to pollinators.

Methods
Our objectives were to study how attractiveness traits of insect-pollinated plants are affected by the presence of neighbouring wind-pollinated plant species. To do so, we set up a greenhouse experiment in which we grew seven plant species in binary mixtures in pots.
Experimental set-up. In March 2012, seedlings of all species were planted in plastic pots (14 cm Ø; 1.5L, Puteaux SA, France) in sandy soil (pH = 6). The soil was taken from a grassland site (CEREEP-Ecotron Ile-de-France, St Pierre-lès-Nemours, France) and was sieved (< 4 mm) to remove rocks and plant material. Six plant individuals were placed in each pot to form two-species mixtures with three individuals of one insect-pollinated species in alternation with three individuals of one wind-pollinated species. We also set up control monocultures with six individuals of the same species (insect-pollinated or wind-pollinated). Each mixture was replicated 5 times, making a total of 95 pots (5 × 3 monocultures of insect-pollinated species, 5 × 4 monocultures of wind-pollinated species, 5 × 4 × 3 binary mixtures). Pots were randomly placed in a greenhouse (CEREEP-Ecotron Ile-de-France, St Pierre-lès-Nemours, France) and their position was changed each week. Plants were watered daily by sub-irrigation (flood floors, DIMAC SAS, France). Air temperature in the greenhouse followed outdoor conditions but was maintained above 18 °C when low temperatures occurred. Photoperiod was initially set at 12-hours per day through natural light and sodium lamps when necessary (i.e. when solar irradiation was under 200 watt/m 2 /hour; HS2000 Hortilux Schréder, The Netherlands). It was adjusted to 16-hours per day to allow for the blooming of L. corniculatus, a long-day flowering species. Because we were mainly interested by belowground competition in this study, we took special care to check that plant foliage did not overlap between individuals all along the experiment. When plant foliage did overlap (especially in mixtures with L. corniculatus) plant supports were put in to separate plant individuals and thus limit aboveground competition (i.e. for light). For all plants, the total number of produced flowers at the end of the flowering period was calculated by summing the daily flower production over the whole flowering period.

Floral traits of insect-pollinated species involved in attractiveness to pollinators.
Nectar traits. For each plant, nectar volume and nectar sugar content were measured on up to three newly opened flowers, after flower size measurements (see above). This ensured that nectar traits were measured on flowers of the same age to limit variations due to flower age 52 . Nectar was sampled using microcapillary tubes (0.5 μ L or 1 μ L; Minicaps end to end, Hirschmann laborgeraete, Germany) and nectar volume was calculated by measuring the length of liquid in the microcapillary tube with a digital caliper (Digit-Cal MK IV, Brown&Sharpe, USA) (μ L.flower −1 .day −1 ). Daily sugar concentration was determined with hand-held refractometers (Eclipse 45-81 and Eclipse 45-82, Bellingham+ Stanley Ltd., UK) calibrated using sucrose solutions (30% and 50% brix). Because nectar not only contains sucrose but also other sugars, our concentration measurements correspond to sucrose equivalent. However, for the sake of brevity, we will only use in the following the term sucrose in reference to "sucrose equivalent". When nectar volumes were too small to be measured by the refractometer (< 0.5 μ L), samples were diluted in Milli-Q water before measurement. If concentration measurements could not be done right after sampling, microcapillary tubes were stored in a refrigerator at 4°C and measured within the next two hours. Because only up to three flowers per plant were sampled, we decided to calculate volume and sucrose indices taking into account the number of flowers produced per plant 47  Plant traits. At the end of the flowering period of each insect-pollinated plant species (on the 3 rd of May 2012 for L. purpureum and on the 2 nd of June 2012 for E. plantagineum and L. corniculatus), above-and belowground biomass of all individuals was harvested. Concerning belowground biomass, we took care of separating root systems of each species. Plant biomasses were oven-dried (65°C, 48h) and weighted (g.plant −1 ).

Competitive interactions.
In order to estimate the intensity of competitive interactions between each focal insect-pollinated plant and its wind-pollinated competitors we calculated the log response-ratio (ln RR) as an index of competition 28 . This index is defined as: where P mix is the biomass of a focal plant when grown in mixtures and P control is the biomass of a focal plants in monoculture pots. In order to have a good assessment of the ln RR as well as a variance, ln RR values for each treatment were calculated as means of all possible combinations of each focal plant in a mixture divided by each focal plant in a monoculture. Because three focal plants were present in mixtures, we considered monocultures as 'mixtures' of 3 focal plants with 3 'competitor' plants of the same species. Values of this index are symmetrical around zero with positive values indicating that focals grow better in mixture (i.e. focals are better competitors) and negative values indicating that focals' growth is negatively affected by competitor (i.e. focals are lower competitors). Ln RR values were calculated from aboveground and belowground biomass but only ln RR calculated from total biomass were used to study the effect of competition on final floral traits as it is a better integrator of competition within both compartments.
Data analysis. All statistical analyses were performed using R 3.1 53 . Linear mixed-effects models were fitted to all measured traits (nlme R package 54 ), with the exception of floral display size and total flower production that were fitted to generalized mixed-effect models with Poisson probability distribution and log link function (lme4 R package 55 ). As ln RR values are calculated from final biomass here, this may be relevant to study the response of total floral traits (values summed all along the flowering period for each plant to obtain a total value per plant) to competition but not for daily floral traits as competition can be dynamic along plant lifespan 36 . As a consequence models were fitted with two different explanatory variables: ln RR values calculated from total biomass as a fixed effect for total floral traits (i.e. total flower production, total sucrose index, total volume index) and wind-pollinated species identity as a fixed effect for daily floral traits (i.e. floral display size, flower size, daily sucrose concentration in nectar, daily nectar volume, daily volume index, daily amount of sucrose in nectar, daily sucrose index). The date was also set as a fixed effect for daily floral traits to take into account the effect of plant age. In all models, pots and date (for the repeated measures on plants) were set as random effects. For linear mixed models, data were transformed using log (e.g. floral traits involving nectar volume), square or square root (e.g. flower size, floral traits involving sucrose concentration) transformations, when necessary. Daily data were then analysed through analysis of covariance (ANCOVA). For total data, whose values were summed all along the flowering period for each plant to obtain a total value per plant, analyses of variance (ANOVA) were performed on these total values. When significant differences were detected, post-hoc comparisons were performed (Tukey all-pair comparisons, Holm method for p-value adjustment were used 56 , multcomp R package 57 ). For the date effect or the interaction term, only significant effects are reported. Because generalized mixed-effect models (glmer, floral display size and total flower production) do not provide p-values, pairwise comparisons with Holm method for p-value adjustment were used 56 .