Artemisia ordosica Krosch. (Asteraceae) is a woody species that is widely distributed in the fixed and semi-fixed sand dunes of northwestern China. This species is a deciduous, multi-stemmed, dwarf shrub that has plumose, linearly lobate leaves and a branch system that consists of old brown branches and purple current-year twigs near the soil surface (She et al., 2017). A. ordosica has a deep taproot system that can reach a depth of 1–3 m, but its lateral roots are mainly distributed in the upper soil layer of 0–30 cm, and limited to a range of 0.4 m in diameter from the trunk (Zhang et al., 2008). A. ordosica is a typical shrub species in Mu Us desert, China and is widely used as a nursery plant species for soil restoration in this region (Li et al., 2011). The leaves of this species often start to expand in early April and the shoot biomass reaches its maximum in July, followed by a flowering season from August to late September and leaf abscission in mid-November. The species is wind dispersed by tiny light seeds over distances up to several miles. In our study, seeds of A. ordosica were collected from wild individuals at Shapotou Desert Experimental Station of the Chinese Academy Sciences, Ningxia Province China (104°43’8”E, 37°26’28”N) in October 2016, where the annual mean precipitation is 191 mm and the annual average temperature is 10.0°C. The collected seeds were kept dry in the dark at room temperature until use.

Soils were collected from a semi-arid restoration field near the Shapotou research station, where A. ordosica is the dominant species. On average, soils from this field have a water content of 1.1%, a C/N ratio of 17.0, and contain 0.28 g/kg total N and P, and 5.31 mg/kg available N and 1.95 mg/kg available P. A soil core borer of 5 cm in diameter was used to collect a total of ca. 600 kg of soil from the 0-15 cm layer. After being collected, soils were fully mixed and immediately sieved through a 5 mm mesh and sterilized in an autoclave. A thorough sterilization of soil often requires that the standard mode of autoclaving (121°C and 100 Kpa for 1 hour) is repeated two or three times. To minimize the time for sterilization yet ensuring thorough sterilization, we instead used a slightly stronger mode of autoclaving at 130°C and 200 Kpa for 15 min. Heat sterilization of soil can change soil physico-chemical properties including increases in pH, electrical conductivity, release of macronutrients, and changes in soil organic matter such as increases in the levels of humic acids. These changes may have affected plant interactions with AMF. For instance, increased release of phosphorus may reduce AMF colonization whereas higher levels of humic acids can often be well combined with successful establishment and growth promoting effects of AMF (Nobre et al., 2013; Cozzolino et al., 2016). Sterilized soils were stored at room temperature until use. Funneliformis mosseae was used as the source of AM fungal inoculum in this study. This species forms symbioses with a wide range of plant species, including A. ordosica and was identified as one of the dominant AMF species in the rhizosphere of A. ordosica (Qian and He, 2009). F. mosseae-BGC NM04A was purchased from the Institute of Plant Nutrition and Resource in Beijing Academy of Agriculture and Forestry Sciences, Beijing, China. The strain was originally collected from another semi-arid field site at Ejin Horo Banner, Inner Mongonia, China, ca. 400 km away from the experimental field, with a mean annual precipitation of 346 mm and average annual temperature of 6.3°C.

In total, 128 pots were prepared and each pot was filled with 4 kg sterilized soil. The pots were assigned to eight treatments representing a full-factorial combination of two mycorrhizal treatments (M+/-), two water treatments (W+/-) and two fertilization treatments (F+/-). Each treatment had 16 biological replicates (2M × 2W × 2F × 16 replicates = 128 pots). To create the fertilization treatment (F+), half of randomly selected pots were individually fertilized with 0.0966 g granules of (NH4)₂HPO₄ and thus ca. 23 mg phosphate and 21 mg nitrogen was added to each pot. The other half of the pots did not receive the fertilizer and was used as the non-fertilized treatment (F-). Therefore, following the fertilization treatment there was 10.56 mg/kg available N and 7.70 mg/kg available P in F+ soils, and 5.31 mg/kg available N and 1.95 mg/kg available P in F- soils. Tap water was used to create water treatments: half of the fertilized and non-fertilized pots individually received tap water every other day, adjusting the soil water content to 4.5% (W+). The water content of the other half of the pots was adjusted to 1.5% (W-). The water contents (4.5% vs. 1.5%) were calculated from the estimated highest and lowest percentage of water retention capacity of the soil in our experimental region within a typical growing season. The amount of added water was calculated by weighing the experimental pot and estimating plant weight from a modeled plant growth curve. The mycorrhizal treatment (M+) was created by adding fifteen grams of vital Funneliformis mosseae inoculum contained in granules consisting of hyphae, spores and substrate to half pots of each above-mentioned water or phosphate treatment prior to seedling transplantation. The other half of the pots received sterilized inocula to create the non-mycorrhizal treatment (M-).

Seeds of A. ordosica were surface-sterilized using 0.1% KMnO₄ for 30 min and rinsed twice in distilled water. Sterilized seeds were air-dried and germinated in commercial culture soils (Green Energy Inc., Wuzhong, China) at 20°C and a 16 h photoperiod for 6 weeks. Similar-sized seedlings were selected and individually transplanted into plastic pots (diameter 25cm, 16.5 cm height). All pots were grouped into 16 blocks and each block consists of one replicate from each of the eight treatment combinations. All the pots were fully watered with distilled water in the early 11 days after transplantation to ensure survival of seedlings, and dead plants were immediately replaced. After this early period when all seedlings successfully survived, plants started to receive the different watering regimes to establish the watering treatments (W+/-). Following the watering treatments, height of all plants was measured every two weeks for the next 90 days. Plants were grown in a greenhouse with a 16 h photoperiod. Natural daylight was automatically supplemented with light from 400-W metal halide lamps to keep the light intensity at ca. 2000 μmol m⁻¹s⁻¹ photon flux density. After these measurements, pots from block 1-8 were transplanted to the field to assess herbivore abundance. Those from the other blocks (block 9-16) were harvested to measure plant traits (see below).

The pots from block 1-8 were transferred to an undisturbed field site where A. ordosica populations were naturally distributed. Pots were dug into the soil with their tops leveling the soil surface and arranged in blocks at least 20 m apart according to their original block identity, with a distance of approximately 80 cm between two pots. Plants received no further fertilization or watering treatment and were naturally exposed to herbivorous insects for 4 days. After this period, plants were individually investigated to examine their colonization by herbivorous insects. Adults of the chrysomelid beetle Chrysolina aeruginosa Fald (Coleoptera: Chrysomelidae) showed high colonization whereas no other herbivorous insect species were found. Therefore, the numbers of C. aeruginosa were recorded as a measure of herbivore abundance.

Plant shoot biomass - During the harvest of the greenhouse plants, each plant was separated into roots, stems and leaves. Plant leaves and stems were separately collected and oven dried at 70°C for 72 h to determine their dry weights. Plant roots were gently washed to remove soil particles and treated as described below.

Root architecture - Intact fresh roots of each plant were individually scanned using an Epson Perfection 4990 Photo scanner. The obtained photos were analyzed with WinRHIZO software (Regent Instruments Inc., Quebec, Canada) to estimate specific root length (SRL), root surface area (RSA), total root volume (TRV) and average root diameter (ARD). Specific root length (SRL) was calculated as total root length divided by total root biomass of an individual plant.

Root biomass - A subset of lateral roots was randomly selected from each plant. After recording their fresh weight, they were stored in 70% alcohol to determine root colonization by F. mosseae (see below). For the remaining roots of each plant, we measured both fresh weight and dry weight after oven drying at 70°C for 72 h. The dry-to-fresh weight ratio of these remaining roots was used to calculate the dry weight of the corresponding root subsample that was used for determining AMF colonization of the corresponding plant.

Mycorrhizal infection - Root colonization by F. mosseae was quantified using a gridline intersection method (Bierman and Linderman, 1981). Briefly, each stored fresh root subsample was cut into at least 100 segments of 0.5-1 cm in length. These root segments were cleared in 10% KOH for 10 min at 95°C, and stained using vinegar (5% acetic acid) and 5% black ink (Hero 440, Shanghai, China) for 8 min at 80–90°C. Stained roots were mounted on slides and checked for mycorrhizal structures (arbuscules, vesicles, spores and intercellular hyphae) under a compound microscope (BH-2; Olympus, Tokyo, Japan) at ×40 magnification. Presence of any of these structures was scored at 100 grid intersections, and mycorrhizal colonization rate of an individual plant was quantified as the percentage of intersections with mycorrhizal structures present (Wang et al., 2015).

Element analyses - Dried plant tissues, viz. leaves, stems and roots, were separately ground to a powder and 1 mg of the powder was weighed into tin capsules. The total nitrogen (N) content was measured using an elemental analyzer (Vario EL III, Elementar, Germany). Tissue samples were acid digested using a mixed solution of H₂SO₄ and H₂O₂ in a microwave oven, which continued until the samples were fully dissolved in the solution. The phosphorus (P) content in plant tissues was determined by ICP-OES (Optima 8300, PerkinElmer, USA).

Photosynthetic pigment - Leaf chlorophyll a (Chl-a), chlorophyll b (Chl-b) and carotenoid (Car) contents were determined according to the method described in Lichtenthaler and Welburn (1983) using 80% acetone as extraction solution. The Chl-a, Chl-b and Car contents were measured by spectrophotometry at wavelengths of 663, 645 and 480 nm, respectively (T60U, PG Instruments Ltd, Japan). Leaf proline content was measured using the acidninhydrine method described in Bates (1973). In brief, 0.25 g fresh leaf sample was weighed and homogenized in 5 ml 3% sulfosalicylic acid at 100°C for 10 min before filtration. L-Proline was used as a standard for the colorimetric determination of the filtrated solution at a wavelength of 520 nm.

Plant functional traits, including plant biomass, root traits, and leaf N and P concentrations were analyzed using a general linear mixed model (GLMM) with a maximum-likelihood (ML) iterative algorithm in R version 3.6.1. In this model, AMF inoculation (M+/-), fertilization (F+/-) and water (W+/-) additions as well as their interactions were added as fixed factors, and the identity of the block to which the plants were assigned was added as a random factor. The model was run using the lmer function in the “lme4” package (Bates et al., 2015), and the significance of the test was estimated using the anova function in the “lmerTest” package (Kuznetsova et al., 2017). In the above models, data on leaf biomass and leaf and stem P content were log-transformed, and data on N content, root P content and SRL were square root-transformed prior to the analyses to meet assumptions of normality of the residual distribution and homogeneity of variances in data.

A repeated measures ANOVA (Zar, 1999) was used to analyze data on plant height, in which the treatments M, F and W were included as fixed factors and plant identity as a repeated subject. Data on insect abundance were analyzed using a generalized linear mixed model, following a “poisson” distribution, where M, F and W as well as their interactions were included as fixed factors, and block and plant height at the time of exposure to insect as random factors. A similar model was run without plant height as a random factor. The models were run using the glm function and the significance of the tests was estimated using the Anova function in the “car” package (Fox and Weisberg, 2019). All analyses were performed using R version 4.2.1 (R Core Development Team, 2022).

Overall, AMF colonization was relatively low in mycorrhizal plants (ca. 5% on average). Surprisingly, roots of plants from the non-mycorrhizal treatments were also colonized, but their colonization rate was significantly lower than that of plants from the mycorrhizal treatments (F = 4.85, p = 0.036, Figure S1 ). Root colonization by AMF was not affected by fertilization and water supply, nor by their interaction (all p >0.20).

Plant height – Plant height at the harvest was significantly enhanced by both additional water supply (+10.0%) and by additional fertilizer supply (+4.5%), and their effects acted synergistically (+20.8%) ( Table 1 , Figure 1 , W × P interaction, p < 0.01). Surprisingly, plant height was unaffected by AMF inoculation when plants received additional water, but was reduced by AMF when additional water was not supplied ( Table 1 , Figure 1 , M × W interaction, p < 0.001). The reduction in plant height by AMF under low water conditions tended to be stronger in plants that also did not receive fertilization (-33.5%) than in plants that received the fertilizer treatment (-10.6%).

Plant biomass – Like plant height, total plant biomass was significantly enhanced by additional water (+38.0%), and fertilizer (9.0%) supply, and their joint effects were strongly synergistic (+234%, Figure 2D , Table 1 , W × F interaction, p < 0.001). The same was observed for individual leaf, stem and root biomass ( Table 1 , Figures 2A–C ). Effects of AMF depended on water conditions. Overall, inoculation with AMF reduced leaf, stem and total biomass, but this effect was only observed in plants that did not receive additional water supply ( Table 1 , M × W, p < 0.05, Figures 2A, B, D ). Notably, when mycorrhizal effects were tested separately under each of the four environmental conditions, biomass reductions were only observed for plant stems under low fertilizer and soil phosphorus conditions (paired t-test, t=5.32, p < 0.05). Contrary to expectation, the benefits of AMF inoculation were thus not more pronounced at lower soil water and nutrient conditions. Root biomass was not influenced by inoculation with AMF ( Table 1 , Figure 2C ).

Photosynthetic and physiological traits – Leaf chlorophyll b, carotenoid, and proline contents were predominantly affected by water treatment. On average, proline content was higher but chlorophyll b and carotenoid contents were lower in leaves of plants that received additional water, whereas leaf chlorophyll a content was not affected by water treatment ( Table 1 , all p<0.01, Figures 3A–D ). The reduction in leaf chlorophyll b content by additional water occurred only in plants that also received fertilization ( Table 1 , F × W, p<0.05, Figure 3B ). When additional fertilizer was supplied, leaf proline content was enhanced by AMF under low water conditions, but reduced by AMF under high water conditions (AMF × F × W interaction, P < 0.05, Table 1 , Figure 3D ).

Root traits – Among the measured plant root traits, SRL was decreased but ARD was increased by fertilization ( Table 1 , Figures 4A, B ). All measured root traits were significantly affected by water addition. But whereas water addition reduced SRL, it enhanced ARD, RSA and TRV. Effects of water addition were stronger in plants that were not fertilized than in fertilized plants (W × F interactions, all p < 0.01, Table 1 , Figures 4A–D ). SRL was enhanced by AMF, but this only occurred when plants did not receive fertilizer supply ( Table 1 , Figure 4A ). The other root traits, including ARD, RSA and TRV were not affected by inoculation with AMF ( Table 1 , Figures 4A–D ) except for ARD that was reduced by mycorrhizal inoculation under low fertilization and water conditions but enhanced by inoculation under high fertilization and water conditions ( Table 1 , Figure 4B ).

Leaf nutritional traits – Concentrations of nitrogen in leaves, stems and roots were strongly enhanced by soil fertilization but reduced by water addition ( Table 2 , Figures 5A–C ). Especially in roots, effects of fertilization on tissue N concentrations were stronger under low water conditions than under additional water supply (W × F interaction, p < 0.05, Table 2 , Figures 5A–C ). For the concentrations of phosphorus in leaves, stems and roots a similar pattern was observed. These were significantly enhanced by fertilization ( Table 2 , Figures 5D–F ), but in leaves and roots this effect was more strongly observed if plants did not receive additional water (W × F interactions, p < 0.05, Table 2 , Figures 5D, F ). Surprisingly, leaf P concentration was overall significantly reduced by AMF (p < 0.01, Table 2 , Figure 5D ). Although the interaction between AMF and fertilizer treatment was not significant, the suppressive effect of AMF on leaf P tended to be stronger under low fertilization conditions (F-W-: t-test, t _{[1, 3]} = 4.89, p < 0.05; F-W+: t-test, t _{[1, 3]} = 2.42, p < 0.10) than under high fertilizer conditions (F+W- and F+W+: both p > 0.6).

Abundance of the herbivore C. aeruginosa was significantly lower on plants inoculated by AMF than on non-mycorrhizal plants, but this effect was only significant and stronger on non-fertilized plants (mean ± s.e. in insect abundance; AMF: 0.19 ± 0.14; non-AMF: 1.31 ± 0.38; t _{[1, 15]} = 2.70, p < 0.05) than on fertilized plants (AMF: 1.56 ± 0.57; non-AMF: 2.47 ± 0.87; t _{[1, 15]} = 1.10, p = 0.29), resulting in a significant two-way interactions between AMF and fertilization treatment ( Table 2 , M × F: χ² = 4.00, p = 0.045, Figure 6 ). In contrast, herbivore abundance was overall higher on plants that had been fertilized, but only when these plants had also received additional watering ( Table 2 , F × W: χ² = 9.44, p = 0.002, Figure 6 ). These treatment effects were not mediated by plant height (a proxy of plant size) at the time of measurement since inclusion of plant height as a covariate in the analyses (χ² = 0.72, p = 0.397) did not alter the significance of the effects of AMF, fertilizer and watering treatments.

Recent years witness an increasing number of studies recognizing that the outcome of plant-AMF interactions often show a continuum ranging from mutualism to parasitism, depending on the context in which the interactions occur (Hoeksema et al., 2010; Johnson et al., 2015; Jin et al., 2017; Qu et al., 2021). However, these results were mostly obtained from studies using short-lived herbaceous plant species and commercial AMF strains as a model system, and it is yet unclear whether these results apply to plants of different life forms. In the current study, a shrub species, A. ordosica was used to measure its responses to AMF under contrasting soil water and soil fertilization conditions. Our results show that inoculation of A. ordosica with a strain of the AM fungus F. mosseae that was isolated from a comparable sandy, low precipitation habitat as the collection site of the host plant seeds, overall reduced plant height, and the reduction consistently increased over time as plants grew. This result may relate to the overall low root colonization (5% on average) by the mycorrhizal strain, indicating a potentially limited opportunity for beneficial C-P trade in this symbiotic combination (Graham and Abbott, 2000). Such low colonization rate may result in low transfer of water and nutrients to the plant. If, even at low levels of colonization, the AMF would still suppress the plant’s direct pathway of P uptake through its own root system, this cost could outweigh the benefits of the indirect pathway provided by mycorrhizal extra-radical mycelia (Smith et al., 2011) and jeopardize the beneficial C-P trade traditionally hypothesized for mycorrhizal symbiosis (Grace et al., 2009; Johnson, 2010). The reason for the overall low root colonization of A. ordosica in our experiment is unknown, but low colonization rates of F. mosseae have been observed in other plant species as well. For instance, average colonization percentages of a single F. mosseae strain varied between 2.6 and 27.0% across a range of tomato cultivars (Steinkellner et al., 2012). Interestingly, the cultivar with the lowest colonization percentage nevertheless showed a significant 30% increase in root dry weight in response to inoculation, whereas the root weight of the cultivar with the highest colonization percentage was not affected by F. mosseae. This indicates that, whatever the reason for low colonization is, even low percentages of colonization can significantly affect plant performance, and that the magnitude of AMF effects is not necessarily related to percentage of colonization. It should be noted that despite the low colonization in M+ plants and the unexpected presence of colonization in M- plants, possibly due to cross-contamination, AMF colonization percentages were still significantly higher in M+ plants than in M- plants, indicating that the validity of AMF treatments in our study was not compromised.

Given that F. mosseae was one of the dominant AMF species in the study area and that the F. mosseae isolate used in the current study originated from a region similar to the region where the plant species A. ordosica is commonly found, potential incompatibility between the two partners due to an ecological mismatch should not be a reason for the observed mycorrhizal growth depression (Jin et al., 2017; Řezáčová et al., 2017). Surprisingly, the mycorrhizal plant growth reduction was most strongly observed under low soil water and phosphorus conditions. This is in contrast with other studies showing that AMF inoculation usually alleviates adverse impacts of drought stress (Duc et al., 2018; Liu et al., 2021; Jongen et al., 2022). The mechanism underlying this drought-induced mycorrhizal plant growth reduction is unknown, but it may relate to the observed interactive negative effects of mycorrhizae and drought on root traits, such as specific root length (SRL) or average root diameter (ARD) (Chen et al., 2016). Plants with higher SRL or lower ARD as observed in the current study tend to have greater plasticity in water and nutrient uptake, but they often show less mycorrhizal dependency, and this may be the reason for the mycorrhizal plant growth depression (Eissenstat, 1992). Even more striking was the observation that the mycorrhizal growth reduction tended to be stronger under non-fertilized than under fertilized conditions. Numerous studies have shown that the mycorrhizal growth response becomes more beneficial for plants at lower nutrient levels (e.g. Vogelsang et al., 2006; Klironomos et al., 2011; Johnson et al., 2015), although exceptions have been reported (Püschel et al., 2016; Raya-Hernandez et al., 2020). Perhaps competition for nutrients between AMF and plants occurred under these experimental conditions, which may have aggravated the AMF-induced plant growth reduction (Li et al., 2008).

The mycorrhizal growth reduction was not only observed for plant height but also for plant biomass. Our results showed that inoculation of AMF significantly reduced plant biomass, and, as observed for plant height, the strength of the AMF-induced reduction in plant biomass production depended on soil water and nutrient conditions. Leaf and stem biomass was more strongly suppressed by AMF inoculation under low water or nutrient supply than under more favorable conditions. In addition to the potential competition between AMF and host plants as previously suggested, an alternative explanation for this observation may be that the AMF strain was not adapted to perform optimally with the host under the low water and nutrient conditions since the site of origin of AMF strain used in this study has a higher mean annual precipitation than the experimental region from which the host plants were collected, mimicked by the low water and nutrient conditions. On the other hand, it is interesting to notice that root biomass was not influenced by AMF inoculation. This result indicates that the plant growth response to AMF inoculation was more sensitive to soil heterogeneity in the shoots than in the roots, suggesting the dependence of AMF-host interactions on environmental conditions may differ among plant functional organs (Roesti et al., 2006; Yang et al., 2016).

Many studies have shown that the photosynthetic capacity of mycorrhizal plants is often higher than that of non-mycorrhizal plants (Liu et al., 2015; Zhang et al., 2018; Balestrini et al., 2020). Such improvements can be incurred by enhanced chlorophyll contents, photosynthetic rate and transpiration rate in plants following AMF colonization (Chandrasekaran et al., 2019). However, in our study we did not observe significant effects of AMF inoculation on the concentration of chlorophyll a, chlorophyll b or carotenoids in A. ordosica. This result is in contrast to the results of Zhu et al. (2011) who showed that all these photosynthetic traits were higher in maize colonized by the AMF species Glomus etunicatum than in non-mycorrhizal plants. The lack of AMF effects on photosynthesis-related traits in our experiment may be explained by the extremely low AMF colonization rate of A. ordosica that might be insufficient to systemically induce changes in the plant’s light harvesting capacity (Evelin et al., 2009). AMF inoculation also did not significantly enhance leaf levels of the osmolyte proline. Osmolytes are often synthesized under water stress in order to maintain osmotic balance (Furlan et al., 2020), but this was not observed in our experiment. The reasons for this unexpected result are unknown, but it may be related to the unfavorable growing environment in the greenhouse, e.g. low intensity of light, that can prevent stress-induced plant responses (Lin et al., 2019).

In this study, we exposed AMF inoculated and control plants grown under different soil water and nutrient conditions to natural herbivores. The relatively lower abundance of the herbivore species C. aeruginosa on mycorrhizal than on non-mycorrhizal plants when grown under low fertilizer conditions suggests a reduced attractiveness of AMF-inoculated plants for this insect species. This result complies with the finding that AMF generally induce resistance to leaf chewing herbivores (Pozo and Azcon-Aguilar, 2007; Koricheva et al., 2009; Jung et al., 2012; Meier and Hunter, 2018; Jiang et al., 2021). However, this effect was only observed in plants grown under low soil P, illustrating the context-dependency of the effects of AMF on plant-herbivore interactions. This observation is in line with other studies. For instance, Wang et al. (2020) showed that only in low-nutrient soils AMF colonization was high but aphid infestation was low. Not only soil P content, but also other environmental factors such as light were reported to influence the effects of mycorrhizal inoculation on plant-insect interactions (Qu et al., 2021). Several mechanisms have been proposed to underlie plant defense responses to AMF inoculation, including priming effects or enhanced production of defensive metabolites (Koricheva et al., 2009). In our study, the observed lower incidence of C. aeruginosa on AMF-inoculated plants under low fertilization conditions could have been related to the observed lower P concentrations in mycorrhizal plant leaves under these conditions. Even though AMF incurred only a modest reduction in leaf P under these conditions, overall leaf P levels under these conditions were very low, so that further reductions may have had a significant impact on herbivore preference; lower leaf P concentrations generally represent a lower diet quality for herbivores (Real-Santillan et al., 2019). Interestingly, under higher fertilization levels, the preference of the herbivore for non-mycorrhizal over mycorrhizal plants disappeared. Since the overall levels of leaf phosphorus were much higher under these conditions, this could indicate that once a threshold P concentration of leaves has been reached, the preference of C. aeruginosa is no longer driven by P-demand, but by other stoichiometric resource requirements such as a higher N demand under conditions where P is no longer limiting, or by other factors not significantly affected by mycorrhizal inoculation. Leaf nitrogen is another important leaf trait that affects leaf nutritional quality for herbivores either through provisioning of primary metabolites or through N-based secondary metabolite production, but leaf N in our experiment was not significantly affected by mycorrhizal inoculation. An alternative explanation for the lower abundance of the chrysomelid beetle on AMF-inoculated plants grown under low soil fertilization could be the AMF-induced reduction in leaf and stem biomass under these conditions. However, inclusion of plant height as a covariate in the analysis did not account for any variation in herbivore abundance, so this idea is not supported by our data.