The Interactive Effect of Elevated CO2 and Herbivores on the Nitrogen-Fixing Plant Alnus incana ssp. rugosa

Many studies have found that future predicted CO2 levels can increase plant mass but dilute N content in leaves, impacting antiherbivore compounds. Nitrogen-fixing plants may balance their leaf C:N ratio under elevated CO2, counteracting this dilution effect. However, we know little of how plants respond to herbivores at the higher CO2 levels that occurred when nitrogen-fixing plants first evolved. We grew Alnus incana ssp. rugosa was grown at 400, 800, or 1600 ppm CO2 in soil collected from the field, inoculated with Frankia and exposed to herbivores (Orgyia leucostigma). Elevated CO2 increased nodulated plant biomass and stimulated the nitrogen fixation rate in the early growth stage. However, nitrogen-fixing plants were not able to balance their C:N ratio under elevated CO2 after growing for 19 weeks. When plants were grown at 400 and 1600 ppm CO2, herbivores preferred to feed on leaves of nodulated plants. At 800 ppm CO2, nodulated plants accumulated more total phenolic compounds in response to herbivore damage than plants in the non-Frankia and non-herbivore treatments. Our results suggest that plant leaf defence, not leaf nutritional content, is the dominant driver of herbivory and nitrogen-fixing plants have limited ability to balance C:N ratios at elevated CO2 in natural soil.


Introduction
The growing conditions when major plant evolutionary events took place are quite different from the present. Vascular plants originated and adaptively radiated during the Early Silurian period ca. 440 million years ago [1]. The conditions during this period included an atmospheric CO 2 level of 3300-3600 ppm [2]. The evolution of nitrogen-fixing plants (i.e., the symbiosis with nitrogen fixation bacteria involving the formation of root nodules) occurred during the rapid expansion of flowering plants in the late Cretaceous ca. 100 MYA [3]. During this time, atmospheric CO 2 was ca. 1600 ppm, around four times the present atmospheric level [3][4][5][6]. The nitrogen-fixing symbiosis was likely more advantageous under conditions of ancient CO 2 levels. Many studies have found that future predicted CO 2 levels (<800 ppm) increases plant nitrogen fixation rates [7][8][9]. This increase has been shown to allow plants to maintain their leaf C:N ratio, whereas non nitrogen-fixing plants show increases in leaf C:N ratio with increasing CO 2 levels [10]. However, these studies often use experimental protocols that provide plants with additional nutrients, especially P, which can stimulate an increase in nodule number or mass, but may not represent how nitrogen-fixing plants respond under natural soil conditions. Therefore, it is possible that the decrease in atmospheric CO 2 over geological time restricted the success of existing, and evolution of new nitrogen-fixing species [6,11]. This would also explain the loss of the nitrogen-fixing trait that has occurred within the clade of nitrogen-fixing plants [11][12][13].
Many studies simulating predicted CO 2 level increases in the coming decades have shown that small increases in atmospheric CO 2 increase plant growth. Elevated atmospheric CO 2 levels increase plant water use, allowing for increased carbon assimilation [14].

Plant Biomass
Plants that developed nodules were 55 times larger than plants without (Figure 1). Nitrogen-fixing plants also increased biomass in response to increasing CO 2 , being ca. 1.5 larger at the 800 ppm and 1600 ppm CO 2 level than plants grown at 400 ppm CO 2 . Non-nodulated plants showed no response to increasing CO 2 (F = 51.84, p < 0.0001, for the inoculation treatment by CO 2 treatment interaction). A non-linear (quadratic) model of nodulated plant biomass versus CO 2 level predicted that plant biomass would peak at 1137 ppm CO 2 (F = 39.50, p = 0.003, R 2 = 0.63, Figure S1). Herbivory, by itself or crossed with CO 2 level or Frankia inoculation, did not have a significant effect on plant total biomass. prefer to feed on leaves from nitrogen-fixing plants when given a choice but, 5. herbivores consume more leaf tissue as the C:N ratio increases.

Plant Biomass
Plants that developed nodules were 55 times larger than plants without (Figure 1). Nitrogen-fixing plants also increased biomass in response to increasing CO2, being ca. 1.5 larger at the 800 ppm and 1600 ppm CO2 level than plants grown at 400 ppm CO2. Nonnodulated plants showed no response to increasing CO2 (F = 51.84, p < 0.0001, for the inoculation treatment by CO2 treatment interaction). A non-linear (quadratic) model of nodulated plant biomass versus CO2 level predicted that plant biomass would peak at 1137 ppm CO2 (F = 39.50, p = 0.003, R 2 = 0.63, Figure S1). Herbivory, by itself or crossed with CO2 level or Frankia inoculation, did not have a significant effect on plant total biomass. The effect of CO2 level on Alnus incana ssp. rugosa total biomass with and without nitrogen-fixing nodules. Bars are means with standard errors and averaged across non-herbivore and herbivore treatments. Bars with the same letters are not significantly different according to Tukey's HSD Test.

Leaf C:N Ratio
Frankia inoculation decreased leaf C:N ratio, but only at the 400 ppm CO2 level (F = 3.41, p = 0.04 for the inoculation by CO2 level interaction, Figure 2). At 400 ppm, Frankia inoculated plants had a leaf C:N ratio that was half of that in the non-inoculated treatment. At 800 and 1600 ppm CO2, leaf C:N ratio was high, regardless of the plants being inoculated with Frankia or not. Figure 1. The effect of CO 2 level on Alnus incana ssp. rugosa total biomass with and without nitrogenfixing nodules. Bars are means with standard errors and averaged across non-herbivore and herbivore treatments. Bars with the same letters are not significantly different according to Tukey's HSD Test.

Leaf C:N Ratio
Frankia inoculation decreased leaf C:N ratio, but only at the 400 ppm CO 2 level (F = 3.41, p = 0.04 for the inoculation by CO 2 level interaction, Figure 2). At 400 ppm, Frankia inoculated plants had a leaf C:N ratio that was half of that in the non-inoculated treatment. At 800 and 1600 ppm CO 2 , leaf C:N ratio was high, regardless of the plants being inoculated with Frankia or not.

Nitrogen Fixation
After plants had been fed on and artificially damaged, nitrogenase activity per plant was higher in plants grown at a higher CO2 level, regardless of whether or not they were part of the herbivory treatment (F = 14.73, p < 0.0001), increasing from 14.73 ± 1.01 µmol C2H4 h −1 plant −1 at 400 ppm CO2, to 20.59 ± 1.91 and 27.18 ± 1.41 µmol C2H4 h −1 plant −1 at

Nitrogen Fixation
After plants had been fed on and artificially damaged, nitrogenase activity per plant was higher in plants grown at a higher CO 2 level, regardless of whether or not they were part of the herbivory treatment (F = 14.73, p < 0.0001), increasing from 14.73 ± 1.01 µmol C 2 H 4 h −1 plant −1 at 400 ppm CO 2 , to 20.59 ± 1.91 and 27.18 ± 1.41 µmol C 2 H 4 h −1 plant −1 at 800 and 1600 ppm CO 2 , respectively ( Figure S2). At the time of harvest, there was no effect of CO 2 level (F = 0.84, p = 0.43) or herbivory (F = 0.00, p = 0.99) on rates of nitrogenase activity per plant, but the rate had dropped to 1.8 ± 0.1 µmol C 2 H 4 h −1 plant −1 averaged over all treatments. When nitrogenase activity at the time of harvest was normalized to total plant leaf mass (F = 7.86, p = 0.0009), the rate decreased with increasing CO 2 level (Figure 3), with no effect of herbivore damage (F = 0.28, p = 0.59) or interaction between herbivore damage and CO 2 level (F = 0.81, p = 0.44). This was reflected in patterns of plant mass allocation to nodules, which decreased by 32%, from 2.20 ± 0.08% of total mass at 400 ppm CO 2 to 1.60 ± 0.13 and 1.50 ± 0.07% at 800 and 1600 ppm CO 2 , respectively (F = 23.11, p < 0.0001), with no effect of herbivory (F = 1.97, p = 0.16), or interaction between CO 2 level and herbivory (F = 1.5457, p = 0.2110). Stable isotope analysis indicated that all inoculated plants got most of their nitrogen from fixation, with a mean value of 98.56% ± 0.38%. Consequently, CO 2 levels did not affect the proportion of nitrogen in the inoculated plants derived from fixation (F = 0.27, p = 0.76).

Nitrogen Fixation
After plants had been fed on and artificially damaged, nitrogenase activity per plant was higher in plants grown at a higher CO2 level, regardless of whether or not they were part of the herbivory treatment (F = 14.73, p < 0.0001), increasing from 14.73 ± 1.01 µmol C2H4 h −1 plant −1 at 400 ppm CO2, to 20.59 ± 1.91 and 27.18 ± 1.41 µmol C2H4 h −1 plant −1 at 800 and 1600 ppm CO2, respectively ( Figure S2). At the time of harvest, there was no effect of CO2 level (F = 0.84, p = 0.43) or herbivory (F = 0.00, p = 0.99) on rates of nitrogenase activity per plant, but the rate had dropped to 1.8 ± 0.1 µmol C2H4 h −1 plant −1 averaged over all treatments. When nitrogenase activity at the time of harvest was normalized to total plant leaf mass (F = 7.86, p = 0.0009), the rate decreased with increasing CO2 level ( Figure 3), with no effect of herbivore damage (F = 0.28, p = 0.59) or interaction between herbivore damage and CO2 level (F = 0.81, p = 0.44). This was reflected in patterns of plant mass allocation to nodules, which decreased by 32%, from 2.20 ± 0.08 % of total mass at 400 ppm CO2 to 1.60 ± 0.13 and 1.50 ± 0.07 % at 800 and 1600 ppm CO2, respectively (F = 23.11, p < 0.0001), with no effect of herbivory (F = 1.97, p = 0.16), or interaction between CO2 level and herbivory (F = 1.5457, p = 0.2110). Stable isotope analysis indicated that all inoculated plants got most of their nitrogen from fixation, with a mean value of 98.56% ± 0.38%. Consequently, CO2 levels did not affect the proportion of nitrogen in the inoculated plants derived from fixation (F = 0.27, p = 0.76). The effect of CO2 level and herbivores on nitrogenase acivity per leaf mass at harvest. Blue bars indicated plants without herbivore damage, and red bars indicated with herbivore damage. Different letters indicate significant differences between CO2 levels. There was no difference between herbivore and non-herbivore treatments. Bars are means with standard error. Figure 3. The effect of CO 2 level and herbivores on nitrogenase acivity per leaf mass at harvest. Blue bars indicated plants without herbivore damage, and red bars indicated with herbivore damage. Different letters indicate significant differences between CO 2 levels. There was no difference between herbivore and non-herbivore treatments. Bars are means with standard error.

Leaf Damage and Antiherbivore Compounds
After five days of feeding, 18.24% ± 3.22% of the leaves on non-inoculated plants were damaged by the herbivores, compared to 12.97% ± 1.30% on the nodulated plants (F = 3.49, p = 0.06). Increased CO 2 level had no effect on the proportion of leaves that were herbivore damaged (F = 1.48, p = 0.23).
Elevated CO 2 increased leaf total phenolic concentration by the end of the growth period, increasing by 35% at 1600 ppm CO 2 compared to 400 ppm and 800 ppm CO 2 treatments (F = 27.06, p < 0.0001, Figure 4). Phenolic concentration also varied between herbivory treatments and plants with and without nitrogen-fixing nodules (F= 10.00, p = 0.003, for the interaction effect), but there was no three-way interaction between the treatments (F = 1.98, p = 0.15). We therefore examined the effect of Frankia inoculation and herbivore exposure at each CO 2 level separately. At 800 ppm CO 2 , plants that were both able to fix nitrogen and exposed to herbivores had 18% higher phenolic concentration than nitrogen-fixing plants not exposed to herbivores or non-inoculated plants exposed to herbivores. At the other two CO 2 levels, there was no effect of Frankia inoculation or herbivory exposure on leaf phenolic levels. averaged across all treatments, and specific peroxidase (POD) activity was 11.0 ± 3.0 unit min −1 mg −1 protein. The treatments had no effect on the level of this enzyme activity. Ther was also no relationship between antiherbivore enzyme activities and the degree of leav damage (F = 1.54, p = 0.22). We found no effect of the treatments on PPO activity when expressed on a leaf mass basis.  Specific polyphenol oxidase (PPO) activity was 84.2 ± 21.7 units min −1 mg −1 protein averaged across all treatments, and specific peroxidase (POD) activity was 11.0 ± 3.0 units min −1 mg −1 protein. The treatments had no effect on the level of this enzyme activity. There was also no relationship between antiherbivore enzyme activities and the degree of leave damage (F = 1.54, p = 0.22). We found no effect of the treatments on PPO activity when expressed on a leaf mass basis.

Herbivore Choice Experiment
The leaf consumption area was log transformed due to the large differences in variation between treatments. Averaged across all CO 2 levels, herbivores preferred to feed on nitrogen-fixing plant leaves instead of leaves from non-nodulated plants (F = 74.14, p < 0.0001, for the inoculation effect). However, there was an interactive effect between CO 2 level and Frankia inoculation on leaf consumption (F = 4.80, p = 0.01, Figure 5). Herbivores showed greater preference for leaves from inoculated plants that were grown at either 400 or 1600 ppm CO 2 , but there was no significant difference in the consumption of leaves between inoculated or non-inoculated plants grown at 800 ppm CO 2 . The greatest leaf consumption occurred on leaves from nitrogen-fixing plants raised at 1600 ppm CO 2 . There was no significant difference in the amount of leaf tissue consumed between non-inoculated plant growth at different CO 2 levels. Pooling Frankia inoculated and noninoculated plants together, herbivores consumed almost twice the mass leaf tissues grown at 1600 ppm CO 2 compared with those grown at 800 ppm and 400 ppm (F = 3.25, p = 0.06, for a one-way ANOVA).
consumption occurred on leaves from nitrogen-fixing plants raised at 1600 ppm CO2. There was no significant difference in the amount of leaf tissue consumed between noninoculated plant growth at different CO2 levels. Pooling Frankia inoculated and non-inoculated plants together, herbivores consumed almost twice the mass leaf tissues grown at 1600 ppm CO2 compared with those grown at 800 ppm and 400 ppm (F = 3.25, p = 0.06, for a one-way ANOVA).

Discussion
Many plants benefit from small (industrial era level) increases in atmospheric CO2 by increasing their photosynthetic rate, which boosts growth and yield [25,26]. Without nodules, our plants performed extremely poorly in soil from the field where alders are usually found. Consequently, only nodulated plants showed a growth increase with an increase in atmospheric CO2. The lack of an increase in growth from 800 to 1600 ppm CO2, and predicted peak in growth at around 1100 ppm CO2, is consistent with other findings suggesting present day plants have a growth optimum below ancient levels of CO2 [16,27]. A meta-analysis also concluded that stimulation of plant growth by small CO2 increases is dependent on soil nutrient availability [28]. The authors reported that under high soil nitrogen, elevated CO2 increased aboveground plant growth by an average of 20.1%, while the response to elevated CO2 was only 8.4% under low soil nitrogen availability. Our study suggests that present day plants cannot acclimate the CO2 level under which symbiotic nitrogen fixation evolved, even when they have nitrogen-fixing nodules. Studies using small increases in CO2 have shown that nitrogen fixation rate was increased only when non nitrogen mineral nutrients were supplied [29]. While this suggests that nitrogen-fixing plants could respond positively to ancient CO2 levels if they are also supplied with these nutrients, it is unlikely that high nutrient availability was common in the past.

Discussion
Many plants benefit from small (industrial era level) increases in atmospheric CO 2 by increasing their photosynthetic rate, which boosts growth and yield [25,26]. Without nodules, our plants performed extremely poorly in soil from the field where alders are usually found. Consequently, only nodulated plants showed a growth increase with an increase in atmospheric CO 2 . The lack of an increase in growth from 800 to 1600 ppm CO 2 , and predicted peak in growth at around 1100 ppm CO 2 , is consistent with other findings suggesting present day plants have a growth optimum below ancient levels of CO 2 [16,27]. A meta-analysis also concluded that stimulation of plant growth by small CO 2 increases is dependent on soil nutrient availability [28]. The authors reported that under high soil nitrogen, elevated CO 2 increased aboveground plant growth by an average of 20.1%, while the response to elevated CO 2 was only 8.4% under low soil nitrogen availability. Our study suggests that present day plants cannot acclimate the CO 2 level under which symbiotic nitrogen fixation evolved, even when they have nitrogen-fixing nodules. Studies using small increases in CO 2 have shown that nitrogen fixation rate was increased only when non nitrogen mineral nutrients were supplied [29]. While this suggests that nitrogen-fixing plants could respond positively to ancient CO 2 levels if they are also supplied with these nutrients, it is unlikely that high nutrient availability was common in the past.
In our study, both future and ancient CO 2 levels increased leaf C:N ratio, regardless of the ability of the plants to fix nitrogen. Therefore, the growth conditions in this experiment do not support our first prediction that nitrogen-fixing plants can balance their C:N ratio when CO 2 levels increase to the levels we used. Previous studies have shown that nonnitrogen fixing plants leaf C:N ratio increased under elevated CO 2 (720 ppm) when plants were grown in low (2.2 g N m −2 yr −1 ) or intermediate levels (6.7 g N m −2 yr −1 ) of nitrogen availability [30]. Increased C:N ratios due to high CO 2 levels can lead to increased reactive oxygen species and leaf senescence [31,32]. The inability of nitrogen-fixing plants to maintain their C:N ratio that we found can be attributed to a reduced biomass allocation to nodules and, consequently, a decrease in the amount of nitrogen fixed per leaf mass under elevated CO 2 . There are contradictory reports of the effects of elevated CO 2 on nitrogen fixation and plant nitrogen content [33]. Some studies show small increases in CO 2 (550 ppm) stimulates symbiotic nitrogen fixation [34], with up to a doubling of the rate of nitrogen fixation in black locust (Robinia pseudoacacia) at 700 ppm CO 2 [7]. In this latter study, plants were fertilized weekly with Hoagland-based nutrient solution for over one year, so other essential nutrients were not a limiting factor for nitrogen fixation. However, in natural ecosystems, nitrogen fixation may decrease in the long term with increasing CO 2 due to the limitation of essential elements, especially phosphorus, molybdenum and iron [29,35]. Hungate et al. [36] found that nitrogen fixation at elevated CO 2 in Galactia elliottii Nutt. in scrub-oak vegetation in central coastal Florida increased during the first year but then declined by the third year and subsequent years. This was due to Molybdenum, a required cofactor for nitrogenase, becoming limited under elevated CO 2 . Edwards et al. [37] also found that elevated CO 2 (700 ppm) had no effect on nitrogen fixation in white clover (Trifolium repens) with a low phosphate application (0.04 mM), but increased with a high phosphate application (1.0 mM). The meta-analysis of de Graff et al. [28] found that elevated CO 2 increased nitrogen fixation by 51% when nutrients other than nitrogen were also applied, but had no effect on nitrogen fixation in the absence of nutrient additions. Our study found that whole-plant nitrogen fixation rate increased under elevated CO 2 at week 14-15, but dropped by week 19 regardless of CO 2 levels. As our whole plant rates of nitrogen fixation are a function of plant mass, the early differences in fixation may simply be due to differences in plant mass. The decrease in nitrogen fixation over time suggested increasing nutrient limitation on this process, which is to be expected since our fertilization stopped after 12 weeks. It is possible that the nutrient-poor soil we used prevented nodulated plants from being able to maintain their tissue C:N ratio. It is also likely that in the future, nitrogen-fixing plants will find themselves in conditions with less available nutrients as the increased carbon content of litter from rising CO 2 levels can result in immobilization of plant nutrients [38].
We did find evidence for our second prediction. Carbon-based antiherbivore compounds did increase with increasing CO 2 . Additionally, nitrogen-fixing plants did maintain their levels of N-based antiherbivore compounds, even though the nitrogen-fixing plants were not able to maintain their C:N ratio. We also found that having the ability to fix nitrogen increased total phenols when the plants were exposed to herbivores at 800 ppm CO 2 . This supports other studies showing that nitrogen fixation increases inducible herbivore defence [23,39] and may provide some feeding deterrence at future, but not ancient, CO 2 levels.
Our results also confirmed the third prediction that plants get less leaf damage when leaf tissue accumulates more total phenolic compounds at elevated CO 2 . However, we did not find that the nitrogen-based antiherbivore compound production was related to the degree of plant damage from herbivores. Total phenolic compounds play a major role in host resistance to herbivores [40][41][42] because they can bind to insects' digestive enzymes and interfere with animal digestion [43]. This defence mechanism can be present constitutively or induced after damage by herbivores [39]. Several studies have shown that plants exhibit many inducible defence mechanisms [23,39]. They can both deter feeding and retard insect development [44]. Previous studies have also shown that total phenolic compounds of maize (Zea mays) increased under elevated CO 2 [45]. Previous studies have also shown that plants can accumulate more N-based secondary compounds (e.g., polyphenol oxidase, peroxidase) to defend against herbivore damage [39,46]. These compounds catalyze the oxidation of phenolics to quinones, which bind to leaf protein and inhibit protein digestion in herbivores [39,46]. Our results support the previous study that Alnus exhibits an inducible defence mechanism in response to herbivore feeding by maintaining polyphenol oxidase and peroxidase activity under elevated CO 2 .
The choice experiment showed that herbivores generally prefer to eat the nodulated, rather than non-nodulated plants, confirming our 4th prediction. Herbivores also increased consumption at higher CO 2 , confirming our 5th prediction that herbivores consume more when food quality goes down. However, there was no difference in herbivore preference between nodulated and non-nodulated plant leaves when grown at 800 ppm CO 2 . This may be related to the fact that at this CO 2 concentration, nitrogen-fixing plants had a higher concentration of phenolic compounds when they had previously been fed on, suggesting they can have a greater induced antiherbivore response. This result also supports a previous study suggesting that plant leaf defence, not leaf nutritional content, is the dominant driver of herbivore preference [47].

Plant Growth Condition and Treatments
Seedlings of speckled alder (Alnus incana ssp. rugosa) were raised in growth chambers with 16-h day time/8-h night time (22 • C/18 • C, day/night). The plants were grown in a mix of field soil from a Pinus banksiana forest stand that contains Alnus viridis ssp. crispa, mixed with an equal volume of Turface. A previous study has shown that the soil from the site has an inorganic nitrogen level of 10.2 ± 0.6 mg/kg and an extractable phosphate level of 0.98 ± 0.33 mg/kg [48]. Plants were fertilized with 1/16 N-Rorison nutrient solution containing 0.125 mM Ca(NO 3 ) 2 , 1 mM CaCl 2 , 1 mM K 2 HPO 4 , 1 mM MgSO 4 , 0.0534 mM Fe-EDTA, 0.009 mM MnSO 4 , 0.0045 mM H 3 BO 3 , 0.001 mM Na 2 MoO 4 , 0.0015 mM ZnSO 4 , and 0.0015 mM CuSO 4 [49] once a week for the first 12 weeks. The fertilization was stopped before the choice experiment, described below. Half of the plants were inoculated with the nitrogen-fixing symbiont Frankia, which came from crushed root nodules collected from wild plants. The non-inoculated plants received the same dose of inoculum, which was first autoclaved. Plants were divided among six growth chambers, with two chambers each receiving one of three atmospheric CO 2 levels: 400, 800, and 1600 ppm, representing ambient, future and Cretaceous era atmospheric levels, respectively, following Murray et al. [50]. There were 30 plants per combination of inoculation treatment and CO 2 level. In each chamber, carbon dioxide levels were continuously monitored and controlled with a CO 2 sensor connected to a microcontroller, which controlled a CO 2 injection system [51].

Herbivore Choice Experiment
At week 13 after the start of the CO 2 treatments, a choice experiment was set up to determine whether insects prefer non-nodulated or nodulated plants grown at each CO 2 level. We randomly paired one leaf from an inoculated plant with one leaf from a non-inoculated plant grown at the same CO 2 level. Each pair of leaves were placed into a Petri dish with moistened filter paper and along with one white-marked tussock caterpillar (Orgyia leucostigma) larva. The larvae were raised from eggs obtained from the Great Lakes Forestry Centre (Natural Resources Canada) on their artificial diet at 22 • C on a 12 h light: 12 h dark cycle for one month in room level (ca. 400 ppm) CO 2 . Insects were starved for 48 h before placing them into the Petri dishes. Petri dishes were placed under the same conditions where larvae were raised. Nine inoculated and nine non-inoculated leaves were collected from each chamber, for a total of 18 replicates at each level of CO 2 . Leaf areas were measured before and after 24 h of insect feeding to determine insect preference.

Herbivores Treatment
After the choice experiment, half of the plants in each combination of inoculation treatment and CO 2 level (15 plants) were randomly assigned to an herbivore exposure treatment of O. leucostigma, raised and starved as above. The rest were non-herbivore controls. Since it was not possible to control insect movement, all plants exposed to herbivores were moved to one chamber at each CO 2 level for the feeding period and then moved back after feeding. Each plant was exposed to an average of 5 insects (for a total of 145 larvae in each chamber) for five days without controlling the insect's movement. After the insects were removed, the proportion of leaves showing signs of insect feeding was assessed. Then, half of each plant's leaves were artificially removed to standardize plant damage. The plants were grown for another six weeks after artificial leaf removal and then harvested.

Nitrogen Fixation
The activity of the nitrogen-fixing enzyme, nitrogenase, was measured after leaves were damaged (at week [14][15] and at the time of the harvest (week 20) using acetylene reduction assays [49]. For the pre-harvest assays, the acetylene reduction rate was expressed per plant. At the time of harvest, the acetylene reduction rate was also calculated per total mass and leaf dry mass, to give an index of the amount of nitrogen fixed per mass of photosynthetic tissue. After harvest, dry plant leaf tissues were ground in a ball mill. Leaf carbon and nitrogen content, and their isotopic ratios were measured at the Stable Isotope Facility at the University of California, Davis. The percentage of nitrogen derived from fixation (%Ndfa) was calculated using a modification of the equation following Boddey et al. [52]: where δ 15 N reference is the level of δ 15 N in the uninoculated plants, δ 15 N fixing plant is the level of δ 15 N in the inoculated plants (Alnus incana ssp. rugosa) and B is the δ 15 N of nitrogen fixing plants that are fully dependent upon symbiotic nitrogen fixation and sampled at the same growth stage.

C-Based Antiherbivore Compounds
The C-based antiherbivore compounds were analyzed at the time of harvest. Total phenolic compounds were determined using the Folin-Ciocalteu method [53]. In brief, 0.01 g dry leaf powder was incubated in the dark for 24 h in 10 mL of 40% ethanol. After centrifuging at 5000 RPM for 10 min, 1 mL of supernatant was mixed with 0.5 mL 50% Folin-Ciocalteu reagent, incubated for 3 min and then incubated in the dark for 30 min with 1 mL of 5% Na 2 CO 3 . The absorbance was then measured at 750 nm and compared to a standard curve prepared with known concentrations of gallic acid.

N-Based Antiherbivore Compounds
The N-based antiherbivore compounds were analyzed after leaf damage. The polyphenol oxidase and guaiacol peroxidase assays were only performed on Frankia inoculated plants since the non-inoculated plants did not produce enough tissue to conduct all assays. Polyphenol oxidase and guaiacol peroxidase activity were measured relative to the protein content of leaves. Leaf protein content was determined with a Bio-Rad protein assay using the Bradford [54] method. In brief, 0.6 g of frozen leaf tissue was ground with liquid nitrogen and then homogenized in 10 mL extraction buffer consisting of 50 mM K-phosphate (pH 7.1), 1% PVP (polyvinyl-pyrrolidone), 1 mM EDTA (Ethylenediaminetetraacetic acid), and 5 mM ascorbic acid. The homogenate was centrifuged at 23,000× g for 20 min at 4 • C. The supernatant was used for both the protein and antiherbivore compound assays. The protein content was measured after adding the Bradford reagent, using BSA as a standard. Polyphenol oxidase and guaiacol peroxidase were measured following Tscharntke et al. [23]. For the polyphenol oxidase assay, 0.15 mL of supernatant was mixed with 1.1 mL of 50 mM potassium phosphate (pH 7.1) and 0.3 mL of 100 mM catechol. After two minutes, the absorbance at 420 was measured. One unit of polyphenol oxidase was defined as the amount of enzyme that caused an increase in the absorbance of 0.01 per minute, and specific activity was expressed as units min −1 mg protein −1 [55]. The guaiacol peroxidase assay was performed by mixing 0.15 mL of supernatant with 1.2 mL of 50 mM K-phosphate (pH 7.1), 0.15 mL of 20 mM guaiacol, and 0.06 mL of 12.3 mM H 2 O 2 . The enzyme activity was calculated from the increase in the absorbance at 470 nm after 2 min [23]. One unit of activity was defined as the amount of enzyme that increases the optical density 470 by 0.01 per min. The specific enzyme activity was expressed as a change in optical density (∆OD) per min per mg protein (units min −1 mg −1 protein).

Statistical Analysis
Data were analyzed using JMP Pro 14. The data were analyzed using three-factor (CO 2 level, Frankia inoculation, and herbivore exposure) ANOVA models followed by Tukey's HSD tests, with growth chamber nested within CO 2 treatments. The non-inoculation treatment was dropped from the model for the nitrogen fixation response variables, as non-inoculated plants did not fix nitrogen. We also explored the relationships between variables using linear and non-linear least squares models. Since the effect of the CO 2 level on a number of plant responses was not linear, quadratic models were also used to fit the data. Residual plots were used to determine if there are large differences in variation between treatments. If so, the data were log-transformed.

Conclusions
Our study found that nitrogen fixation cannot meet the higher nitrogen demand of plants when grown under 800 and 1600 ppm CO 2 , the former being a CO 2 concentration that may occur in the future and the latter a level that occurred in the distant past. The increased leaf C:N ratio under elevated CO 2 is associated with higher total leaf phenol content. Like non-nitrogen fixing plants, nitrogen-fixing plants at present have likely evolved not to be able to acclimate ancient CO 2 levels. Future studies are needed to test the effect of ancient CO 2 on nitrogen fixation when other nutrients are not limited. Finally, field studies are required to examine the complex relationship between elevated CO 2 , nitrogen fixation and herbivory.
Supplementary Materials: The following are available online at https://www.mdpi.com/2223-774 7/10/3/440/s1, Figure S1: The effect of elevated CO 2 on total biomass of nodulated/non-nodulated Alnus incana ssp. rugosa. (p = 0.027, F = 14.81 from quadratic fit). Non-linear quadratic fit line: Y = 15.99 + 0.017*X − 4.16*10 −5 *(X − 933.33) 2 ; Figure S2: Effect of elevated CO 2 and herbivores on nitrogenase activity per plant after herbivore and artificial damage (at week 14 and 15). Blue bars indicated plants without herbivore damage, and red bars plants with herbivore damage. Bars are mean value with standard error (SE). Letters were from Tukey HSD Test. Different letters indicated a significant difference between CO 2 treatments. There was no significant difference between herbivore and non-herbivore treatments; Figure S3: By the end of week 10, nodulated plants' leaves started to show dark spots, indicative of K deficiency (a,c). At week 10, under elevated CO 2 (800 ppm and 1600 ppm), nodulated plants' leaves margin started to show reddish colour (b). Plant leaf senescence was more severe at 1600 ppm compared with 800 ppm treatments (d).
Author Contributions: Both authors conceived the study and wrote the manuscript. H.C. carried out the experiment and analyzed the data. Both authors have read and agreed to the published version of the manuscript.