Sting nematodes modify metabolomic profiles of host plants

Plant-parasitic nematodes are devastating pathogens of many important agricultural crops. They have been successful in large part due to their ability to modify host plant metabolomes to their benefit. Both root-knot and cyst nematodes are endoparasites that have co-evolved to modify host plants to create sophisticated feeding cells and suppress plant defenses. In contrast, the ability of migratory ectoparasitic nematodes to modify host plants is unknown. Based on global metabolomic profiling of sting nematodes in African bermudagrass, ectoparasites can modify the global metabolome of host plants. Specifically, sting nematodes suppress amino acids in susceptible cultivars. Upregulation of compounds linked to plant defense have negative impacts on sting nematode population densities. Pipecolic acid, linked to systemic acquired resistance induction, seems to play a large role in protecting tolerant cultivars from sting nematode feeding and could be targeted in breeding programs.

ensconced in plant tissue, as is the case with root-knot nematodes, they have choices; if a plant is particularly well defended, they can leave.
Because their ectoparasitic lifestyle gives some plant-parasitic nematodes more options, it is unclear the extent to which co-evolution has led to ectoparasitic nematodes developing mechanisms to hijack and modify the metabolome of host plants to their benefit.
To explore the extent to which ectoparasitic nematodes can modify host plant metabolomes, we used the sting nematode (Belonolaimus longicaudatus Rau) on African bermudagrass (Cynodon transvaalensis Burtt-Davy) as a model system. B. longicaudatus affects a variety of agricultural crops, including peanuts and cotton 19,20 , and is a primary and devastating pathogen of turfgrass 21,22 where it leads to reduced drought tolerance, environmental concerns due to leaching, and, in combination with other stresses, plant death 23 . Effective and environmentally friendly control methods for working with this pathogen have focused on screening and development of tolerant (able to withstand, but not prevent, nematode feeding) bermudagrass cultivars 23,24 , but mechanisms of tolerance are not understood.

Results
To investigate whether sting nematodes modify metabolite production in their host plants, mixed age Belonolaimus longicaudatus nematodes were introduced to the root zones of three African bermudagrass (Cynodon transvaalensis) lines of differing tolerance: one susceptible line (AB03), one moderately tolerant line (AB33), and one tolerant line (AB39). Tolerance in these bermudagrass lines had been determined in previous work; lines are considered tolerant if there was no reduction in root length or had greater root length than the ï¿½Ăï¿½Tifway' cultivar despite B. longicaudatus infection 23,24 . In our work, comparison control plants received no nematodes. Ninety days after inoculation, Plants were unearthed and sting nematode damage assessed. Following damage assessment, untargeted metabolomic profiling was conducted on root samples. Bermudagrass response to sting nematode feeding. Sting nematode population densities were sig- Presence of sting nematode caused significant ( = df 36, = − . t 9 5, < . p 0 0001) reductions in root biomass (Fig. 1C). Despite supporting the highest sting nematode populations, the moderately tolerant line (AB33) showed moderate levels of root biomass loss (Fig. 1D). Even though the susceptible line (AB03) supported lower sting nematode populations, it had significantly higher root biomass loss compared to either the moderate (AB33, = . p 0 003) or the tolerant (AB39, < .
p 0 001). Modifications to the global metabolome are associated with specific compounds. Indicator species analysis (multi-level pattern analysis) was used to calculate the association of each identified (5% of detected compounds) compounds with line and treatment. L-Pipecolic acid was strongly associated (ϕ = .
p 0 001) with plants inoculated with sting nematode. Abundance of known metabolites across lines was grouped using heirarchical cluster analysis then coupled with association values by line (Fig. 3). Guanine, oxoproline, inosine, and citrulline were closely related and highly associated (ϕ > . 0 51, < . p 0 003) with the susceptible AB03 line. A number of compounds were closely associated with the moderately tolerant line AB33 including the closely grouped Orthophosphate, ferrulate, and malate (ϕ > . 0 55, < . p 0 001). Adenine, sugar alcohols, D-glucaronic acid, asparagine, and theophylline were closely associated (ϕ > . compounds responsible for tolerance to sting nematodes. To further examine compounds responsible for differences in observed plant responses to sting nematode infection, a metabolome-wide association approach was taken using a series of Wilcoxon sign-rank contrasts between plants with and without nematodes by line and compound (with correction for the false discovery rate). Among the compounds detected by this approach, amino acid related compounds seemed to play a large role (Fig. 4). In the most susceptible line (AB03), amino acid abundance was significantly ( < . P 0 05) suppressed in seven out of the 13 amino acid compounds assayed. Four of the remaining compounds (L-isoleucine, L-methionione, citrulline, and n-alpha acetyl-L-lysine) showed similar, albeit non-significant, trends. This pattern was not apparent in either the moderately susceptible (AB33) nor tolerant (AB39) lines.
In addition to sting nematodes modifying amino acid production in susceptible bermudagrass plants, production of a number of defence-related compounds was associated with nematode abundance. Higher normalized levels of D-glucuronic acid, glycolate, and phenylalanine were significantly ( = . . . . p 0 035, 0 049, 0 038 respectively) associated with lower levels of sting nematode ( Fig. 5A-C). Although lines had differing levels of phenylalanine and nematodes, the negative relationship (slope of − 89.6, = − . t 2 1, = . p 0 05) was apparent within lines (Fig. 5D).
Differences in L-pipecolic acid were also strongly associated with sting nematode presence. L-pipecolic acid production significantly ( = df 12, = . t 2 74, = . p 0 05 after conservative bonferroni correction) increased in the presence of sting nematode in the tolerant (AB39) line (Fig. 6A). Although there were no within-line trends, lines (particularly the tolerant AB39) with increased production had substantially reduced sting nematode populations (Fig. 6B). Although not significant at α = .
0 05, the susceptible line followed trends of increased L-pipecolic acid and reduced nematode population (Fig. 6A,B).     (Fig. 2). The global metabolome of all three cultivars were substantially different and sting nematode feeding significantly modified metabolome profiles of the susceptible (AB03) and moderately tolerant (AB33) lines. These results from global, untargeted metabolomic profiling suggest that feeding by an nematode ectoparasite can influence and potentially manipulate host plant response similar to endoparasitic nematodes. Similar to altered profiles induced by endoparasitic root-knot and cyst nematodes, sugars and amino acids were associated with patterns of tolerance (Fig. 3).
In particular, sting nematode feeding seemed to have a large effect on amino acid levels in the susceptible line (AB03; Fig. 4). Suppressed levels of amino acids in the susceptible line could be due to a number of factors ranging from active suppression by sting nematode effectors to reduced production as a result of reduced nutrient uptake from a compromised root system (Fig. 1). Feeding by most endoparasitic nematodes (eg root-knot and cyst nematodes) stimulates increased levels of amino acids. In this case, feeding by the ectoparasitic sting nematode is severely compromising host plant health of susceptible plants.
The role of plant defense in mediating host-pathogen interactions. Plant defense related compounds were strongly associated with sting nematode infection across lines (Fig. 5. Increased levels of D-glucuronic acid, glycolate, and phenylalanine were associated with reduced populations of sting nematode; individual plants with high levels of those compounds tended to have lower levels of sting nematode infection. These compounds play important roles in physical and chemical defense of plant tissues. D-glucuronic acid can be a minor contributor to ascorbic acid synthesis, but plays a large role in synthesis of cell wall precursors 26 . The association of higher levels of this cell-wall precursor with lower nematode abundances could suggest that host plants with better physically protected cell walls can better prevent nematode ectoparasite feeding. Glycolate has a major role in photosynthesis 27 ; higher levels of this metabolite associated with reduced sting nematode numbers could indicate that more metabolically active plants are better able to deal sting nematode infection. Phenylalanine can be a precursor for salicylic acid, critically important in plant defense and resistance against pathogens [28][29][30] . Although different lines have different overall levels of phenylalanine, increasing levels of phenylalanine are associated with lower levels of sting nematodes across lines which could indicate a role for the salicylic acid pathway in controlling sting nematode infection (Fig. 5D).
This idea is further bolstered by observed differences in Pipecolic acid production (Fig. 6). Pipecolic acid induces systemic acquired resistance both in conjunction with the salicylic acid pathway, and independently 31,32 . In bermudagrass lines able to maintain lower sting nematode populations, pipecolic acid is increased. In particular, the tolerant (AB39) line with low sting nematode populations and low root biomass loss had significantly higher levels of pipecolic acid following nematode infection (Fig. 6).
preventing sting nematode infection. Specific changes in plant defenses and amino acid production coupled with broad changes in the global metabolome as a result of sting nematode infection provide evidence for ectoparasite modification of host metabolomes. While sting nematodes do not create specialized feeding sites, they nevertheless seem to have co-evolved to the extent that they are able to suppress amino acid production in susceptible cultivars and to induce defense pathways. In a similar situation to root-knot and cyst nematodes, salicylic acid pathways may hold the key to developing tolerance and even resistance against the sting nematode. Root-knot nematodes cause upregulation of phenylalanine and salicylic acid limits root-knot nematode invasion 16 . Similarly, salicylic acid plays a large role in regulating resistant and susceptible responses to cyst nematode 33,34 . In bermudagrass, high levels of compounds related to the salicylic acid pathway and systemic acquired resistance are linked to reduced sting nematode feeding. It seems likely that these compounds are related to mechanisms of tolerance to be focused on in plant breeding. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods organisms. Three African bermudagrass (Cynodon transvaalensis Burtt-Davy) lines were used to evaluate metabolomic responses to pathogen infection. One susceptible line (AB03), one tolerant line (AB39), and one line with moderate tolerance (AB33) were stolon propagated and allowed to establish for 30 days in 3.8 cm diameter, 21cm deep UV-stabilized Ray Leach Cone-tainers (SC10; Stuewe & Sons, Inc., Tangent, OR) filled with 100ml USGA grade sand followed with a soil plug (to prevent sand leakage). Plants were grown in a climate controlled greenhouse at 25 o C and 50% RH under a 11:13 light:dark cycle. Plants were watered twice daily for 5 minutes and received 24-8-16 NPK liquid fertilizer on a weekly basis. Sting nematode (Belonolaimus longicaudatus Rau) originally collected from Florida turf and reared on C. transvaalensis were used in bioassays.

Bioassays.
To evaluate the effects of nematode infection on bermudagrass metabolite response, treated plants in Cone-tainers were inoculated with 50 sting nematodes of mixed age and gender per Cone-tainer. Control (noninoculated) plants did not receive any nematode treatment. Seven replications of each line (AB03, AB33, AB39) and treatment (inoculated, noninoculated) combination were conducted. Ninety days following nematode inoculation (time enough for nematode growth and reproduction), plants were removed from the container, the soil plug removed, and the 100ml of sand gently rinsed to extract the nematodes for further processing through centrifugal flotation and counting on an inverted microscope. Rinsed roots were placed in 50 ml falcon tubes then immersed in liquid nitrogen and stored at − 80 o C until lyophilization. Lyophilzed roots were weighed and samples selected for metabolomics analysis.
Metabolomics. Following weighing the entire root biomass, 0.5 root samples were transferred to 2 centrifuge tubes and ground in a Geno/Grinder 2010 tissue homogenizer with ball bearings. After grinding, metabolites were extracted through addition of 1.5 of 1:1 Methanol:Ammonium Acetate, addition of 20 internal standard mix, vortexing, and centrifugation at 17,000G for 10 minutes. Following centrifugation, 800 of supernatant was transferred to an LC vial and 1 introduced to a Thermo Scientific Dionex Ultimate 3000 UHPLC using reverse phase chromatography with a ACE Excel 2 C18-PFP (100 × 2.1, 2 at 25 o C and a flow rate of 350/min. Analysis began at 100% 0.1% formic acid in water for three minutes than ramped to 80% acetonitrile for 10 minutes where the ratio was held for an additional 3 minutes giving a total run time of 16 minutes.
Following separation by liquid chromatography, samples were introduced to a Thermo Q-Exactive Orbitrap mass spectrometer. All samples were analyzed using positive and negative heated electrospray ionization with a mass resolution of 35,000 at m/z 200 using polarity switching. Probe temperature was held at 350 o C, spray voltage at 3500 V, capillary temperature at 320 o C, sheath gas at 40, and auxillary gas at 10.

Analysis.
Bioassays. The effects of bermudagrass line, treatment, and their interaction on observed nematode numbers were modeled using linear models and analysis of variance. Residual diagnostics were consulted to ensure conformity to assumptions of normality and homoscedasticity while model significance, likelihood ratios, information criteria, coefficient of determination, and residual examination were used to select the best fit models. Post-hoc comparisons were evaluated using Tukey's method for controlling the family-wise error rate.
Similarly, the effects of bermudagrass line, treatment, nematode count, and their interaction on observed root weights were modeled using linear models and analysis of variance. Residual diagnostics were consulted to ensure conformity of assumptions of normality and homoscedasticity while model significance, likelihood ratios, information criteria, coefficient of determination, and residual examination were used to select the best fit models. Outliers were identified and removed through visual examination of residual diagnostics (including QQ-Plots, Cook's Distance, and Leverage) and mean-shift outlier tests. Post-hoc comparisons were evaluated using Tukey's method for controlling the family-wise error rate.
Root biomass loss estimation was accomplished through non-parametric bootstrapping (with 1000 replications) to estimate the difference in root biomass between bermudagrass plants inoculated with sting nematode and noninoculated plants. Differences between median root loss were evaluated with one-sided permutation tests and adjusted using Bonferroni's method for controlling the family-wise error rate.
Metabolomics. Raw mass spectrometry data were exported and uploaded to Metabolomics Workbench (Study ID ST000353). In preparation for analysis, known compounds with more than one retention time were collapsed into a single known compound. Additionally, contaminants and internal standards were removed from future analysis (Appendix A for list of contaminants and standards removed). Following cleaning, missing data (less than 8.3% per sample) were imputed using a k-nearest neighbors approach ( = k 5) 35 . Following imputation, data were normalized using variance stabilizing normalization to adjust for between-run variations 36 .
To determine whether there were differences between line and nematode treatment, canonical correspondence analysis was applied to the global metabolome (both known and unidentified compounds). Results from the canonical correspondence analysis were further evaluated with permutational analysis of variance with 1000 permutations. Line, treatment, and their interaction were evaluated for their effect on the observed metabolomic profiles. The best fit model was chosen based on permutation statistics (permuted F scores), coefficient of determination, deviance metrics, and goodness of fit metrics.
To examine relationships between labeled metabolites and lines, heirarchical cluster analysis was used to group compounds with similar abundances across lines. Indicator species analysis (multi-level pattern analysis) was then used to explore associations of each labeled compound with lines and treatment using Pearson's Φ coefficient of association as the metric.
To examine differences in abundance of individual labeled compounds, a metabolome wide association study approach was taken where Wilcoxon tests were applied to each compound by line to evaluate differences in compound abundance between noninoculated plants without nematodes and inoculated plants infected by Scientific RepoRtS | (2020) 10:2212 | https://doi.org/10.1038/s41598-020-59062-8 www.nature.com/scientificreports www.nature.com/scientificreports/ nematodes. Resultant p values were corrected for the false discovery rate using the Benjamini and Hochberg method 37 .
To further explore the effect of individual compounds on nematode abundance, compounds of interest from the indicator species analysis and metabolome wide association were evaluated for their relationship to observed nematode population levels. To do so, abundances and nematode numbers were normalized by line to account for differences between genotypes then evaluated with linear models to determine the effect of normalized compound abundance on normalized nematode presence. Model fits were evaluated with information criteria, residual examination, model significance, and coefficient of determination. Differences in L-pipecolic acid production were examined through bootstrapping (with 1000 replications) nematode numbers and pipecolic acid levels.