Both Constitutive and Infection-Responsive Secondary Metabolites Linked to Resistance against Austropuccinia psidii (Myrtle Rust) in Melaleuca quinquenervia

Austropuccinia psidii is a fungal plant pathogen that infects species within the Myrtaceae, causing the disease myrtle rust. Myrtle rust is causing declines in populations within natural and managed ecosystems and is expected to result in species extinctions. Despite this, variation in response to A. psidii exist within some species, from complete susceptibility to resistance that prevents or limits infection by the pathogen. Untargeted metabolomics using Ultra Performance Liquid Chromatography with Ion Mobility followed by analysis using MetaboAnalyst 3.0, was used to explore the chemical defence profiles of resistant, hypersensitive and susceptible phenotypes within Melaleuca quinquenervia during the early stages of A. psidii infection. We were able to identify three separate pools of secondary metabolites: (i) metabolites classified structurally as flavonoids that were naturally higher in the leaves of resistant individuals prior to infection, (ii) organoheterocyclic and carbohydrate-related metabolites that varied with the level of host resistance post-infection, and (iii) metabolites from the terpenoid pathways that were responsive to disease progression regardless of resistance phenotype suggesting that these play a minimal role in disease resistance during the early stages of colonization of this species. Based on the classes of these secondary metabolites, our results provide an improved understanding of key pathways that could be linked more generally to rust resistance with particular application within Melaleuca.


Introduction
The biotrophic plant pathogen Austropuccinia psidii, causal agent of the disease commonly known as myrtle rust, has a broad host range across many species within the Myrtaceae family. In Australasia, Myrtaceae predominate in most ecosystems, with more than 2250 species recorded [1]. They are also found in biological hotspots around the globe [2,3]. As such, A. psidii infection is a threat to many fragile ecosystems characterised by a high rate of endemism [4][5][6]. Further, as many species within this family are important globally due to their value in forestry, the essential oil industry as well as for cultural foods, this rust pathogen poses a significant threat economically [5,7]. To date, 539 Myrtaceae species are known to be vulnerable to A. psidii infection worldwide, including Melaleuca quinquenervia [8]. In 2020, the Australian Department of Agriculture, Water and include cuticular waxes [32] and terpene composition in Eucalyptus sp. and M. quinquenervia [29,31], while increasing terpene concentration in older leaves may be associated with ontogenic resistance [33]. These previous studies are only just beginning to inform our understanding of the mechanisms that have evolved within the Myrtaceae that enable certain individuals/species to resist infection by A. psidii.
The broad range of applications of metabolomics demonstrates that this methodology could be used in assessing mechanisms and identifying novel biomarkers of host resistance to A. psidii. M. quinquenervia represents an ideal host species to identify these biomarkers as there are individuals described previously as susceptible, hypersensitive and resistant to A. psidii infection [34]. By evaluating the response of different M. quinquenervia resistance phenotypes to A. psidii, we aimed to determine if a set of metabolites are involved in innate resistance to the disease and to understand what metabolic pathways are altered by the colonisation process.

Austropuccinia psidii Inoculation of Melaleuca quinquenervia
A total of nine M. quinquenervia individuals of previously defined A. psidii infection phenotype (three each of resistant, hypersensitive and susceptible) were chosen for analysis. Plants were rated according to the disease ranking system proposed previously [14,35] and ranked as resistant (susceptibility rating 1, no visible A. psidii pustules or nectrotic legions), three hypersensitive (susceptibility rating 2, evidence of necrotic lesions but no visible A. psidii pustules) and three susceptible (susceptibility rating 5, large visible pustules) (Figure 1). Each plant was coppiced prior to infection to promote branching and the production of new leaves and subsequently inoculated with A. psidii at the Queensland Department of Agriculture and Fisheries. Spore collection and inoculation were as per Pegg et al. 2018; briefly, urediniospores, collected from infected Syzygium jambos trees, were suspended in a sterile distilled water solution to which the surfactant Tween 20 was added at a rate of two drops per 100 mL. The spore concentration in the suspension was assessed using a haemocytometer adjusted to 1 × 10 5 spores/mL. Plants were inoculated using a fine mist spray (29 kPa pressure), generated by a compressor driven spray gun (Iwata Studio series 1/6 hp.; Gravity spray gun RG3). The upper and lower leaf surfaces of the seedlings were sprayed, ensuring all leaves were coated with a fine mist but run-off of the spore suspension was avoided. Once inoculated, seedlings were placed into a Controlled Environment Room (CER) set at 18 • C in the dark for 24 h.
Samples of leaves were collected prior to inoculation (i.e., 0 h), 24 h, 48 h post inoculation (hpi) and 5 days post inoculation (dpi). Six replicates of young leaves of the same plastochron index (to avoid ontogenic effects) from fresh growth on each plant were collected at each time point, weighed and stored at −80 • C prior to analysis.

Leaf Metabolite Extraction
The frozen leaf samples were weighed and immediately ground into powder by beadbeating in 200 µL extraction solvent (4:4:2 methanol:acetonitrile:deionised water) twice for 30 s each. The leaf:solvent mixture was then combined with 300 µL additional extraction solvent, and then vortexed to mix for 10 s. The leaf:solvent mixture was then placed in icy water and sonicated in a sonicating water bath for 25 min. Solid cellular material was removed by centrifugation at 18,000× g for 10 min. The supernatants were collected while avoiding undissolved particles and filtered through a 0.22 µm syringe filter. The resulting metabolite extracts were stored at −80 • C until metabolite profiling was performed.  The represents the adaxial view of the leaves while the bottom row represents the abaxial vie left to right, leaf 1 represents rating (R) 1, the resistant phenotype, displaying no visible pustules or necrotic legions. Leaf 2 represents R2, the hypersensitive phenotype, displaying legions but no visible pustules. Leaves 3-5 represent susceptibility R 3-5, displaying uredi pustules of increasing size. Plants demonstrating susceptibility R5 were selected for the su ity phenotype in this study.

Leaf Metabolite Extraction
The frozen leaf samples were weighed and immediately ground into pow bead-beating in 200 μL extraction solvent (4:4:2 methanol:acetonitrile:deionised twice for 30 s each. The leaf:solvent mixture was then combined with 300 μL ad extraction solvent, and then vortexed to mix for 10 s. The leaf:solvent mixture w placed in icy water and sonicated in a sonicating water bath for 25 min. Solid material was removed by centrifugation at 18,000× g for 10 min. The supernatan collected while avoiding undissolved particles and filtered through a 0.22 μm syr ter. The resulting metabolite extracts were stored at −80 °C until metabolite profil performed.

Ultra Performance Liquid Chromatography High Definition Mass Spectrometry wit Mobility (UPLC HDMS E ) Analysis
Leaf extracts were analysed on a Waters Acquity I-Class UPLC system and a From left to right, leaf 1 represents rating (R) 1, the resistant phenotype, displaying no visible A. psidii pustules or necrotic legions. Leaf 2 represents R2, the hypersensitive phenotype, displaying necrotic legions but no visible pustules. Leaves 3-5 represent susceptibility R 3-5, displaying urediniospore pustules of increasing size. Plants demonstrating susceptibility R5 were selected for the susceptibility phenotype in this study.

Ultra Performance Liquid Chromatography High Definition Mass Spectrometry with Ion Mobility (UPLC HDMS E ) Analysis
Leaf extracts were analysed on a Waters Acquity I-Class UPLC system and a Waters Synapt G2-Si HDMS with a Waters UniSpray Ionisation source at the Western Sydney University Mass Spectrometry facility. The metabolites were separated on a Waters AC-QUITY UPLC HSS T3 1.8 µm 100 × 2.1 mm Column at 35 • C. The injection volume was 2 µL. The mobile phases were A (Water + 0.1% Formic Acid) and B (Acetonitrile + 0.1% Formic Acid). The chromatographic flow rate was 0.5 mL/min with a 9 min gradient, with mobile phase A held at 99% for 1 min, decreased to 85% over 1 min, decreased to 50% over 2 min, decreased to 5% over 2 min and increased to 99% over 2 min. Leucine Enkephalin Lockspray solution (Waters, 1 ng/mL) was used as a standard.
Data acquisition was performed with ion mobility separation followed by mass fragmentation and high-resolution mass analysis. The mass range of metabolites acquired was 50-1200 m/z, the scan time was 0.2 s and the elevated energy transfer collision voltage was 20-50 eV. For this experiment, the instrument was run in positive ionisation mode with the following settings: Capillary: 0.5 kV, source temperature: 120 • C, sampling cone: 30 V,

Statistical and Molecular Analysis
The compound measurement table containing the feature spectral information and peak intensity of all samples (including quality-control samples and solvent blanks) was exported from the Progenesis QI for metabolomics (version 2.4) and further analyses were performed using R (version 3.6.3). The data table was scaled by leaf weight and low intensity peaks (intensity < 500) were removed after blank subtraction. The data table was then imported into MetaboAnalystR (version 3.0.3) [36]. To improve the power of downstream statistical testing, features that are missing in >50% samples or near-constant across all samples were filtered based on interquartile range. Additionally, outliers were removed after visual identification using principal component analysis (PCA). The data table was log-transformed before two-way analysis of variance (ANOVA) with multiple comparisons correction (false discovery rate). The significant features with p < 0.05 were selected for further clustering analyses. Hierarchical clustering and k-means clustering were performed using R (version 3.6.3). The log-transformed data table was also used for permutational multivariate analysis of variance (PERMANOVA) with the pairwise Adonis wrapper (https://github.com/pmartinezarbizu/pairwiseAdonis accessed on 10 December 2021) based on the 'adonis' function of the vegan package (version 2.5-6) on R. Progenesis QI was used to structurally classify the significant metabolites identified by the two-way ANOVA and clustering analyses. Based on the identities provided by the Chemical Entities of Biological Interest (ChEBI), metabolite enrichment analysis was performed on the Metabo-Analyst web platform (https://www.metaboanalyst.ca/ accessed on 10 December 2021). Metabolomics data will be made publicly available following acceptance of the manuscript for publication.

Metabolomic Response to Myrtle Rust Infection in M. quinquenervia
M. quinquenervia plants belonging to one of three phenotypes (resistant, hypersensitive and susceptible) were analysed for their metabolomic responses to the early stages of A. psidii infection. Six leaves of the same plastochron index were sampled prior to inoculation (0 h) and 24 h, 48 h and 5 days after inoculation (total 216 leaves). The metabolome of the leaves at each time point was analysed by LC-MS and after the removal of outliers, five biological replicates were analysed and a total of 11,276 molecular features (MFs) were identified. Analysis of the leaf metabolomes identified that the resistance phenotype of the individual, time post-inoculation and the interaction of these two factors were all significant as determined by permutational multivariate analysis of variance (PERMANOVA; Table 1). Amongst these MFs, we found 2383 MFs have significant difference in abundance under influence of time post infection, phenotype of the plants or the interaction of both (p < 0.05; Top 500 MFs abundance plotted in Figure 2A; Supplemental Table S1). The abundance of 978 and 2157 MFs were influenced by the phenotype of the plants and the time after inoculation, respectively, while 909 MFs were influenced by the interaction of phenotype by time post infection ( Figure 2B).

Organoheterocyclic Compounds Significantly Enriched According to Resistance Phenotype Regardless of Time Post Infection
We further analysed the metabolomics profile of resistant, hypersensitive and susceptible phenotypes of M. quinquenervia in response to infection. Although hierarchical clustering did not clearly separate MF profiles of leaves following infection from different phenotypes, both PERMANOVA and ANOVA results support that there is significant variation in the leaf MF profiles from M. quinquenervia of different resistance phenotypes. In an effort to identify the MFs that distinguish susceptibility of M. quinquenervia to A. psidii, enrichment analysis was performed on the metabolite responses that were significantly influenced by phenotype, irrespective of time (ANOVA adjusted-p < 0.05). Based on the MFs for which structural predictions were made by Progenesis using the CHEHBI database, organoheterocyclic compounds, nucleic acids, organic acids and carbohydrates appeared to be the most enriched metabolite sets that differentiated these phenotypes following application of the pathogen ( Figure 2C; Supplemental Table S2).

Few Metabolites Are Significantly Associated in a Resistance Phenotype × Time Post-Infection Manner
To establish the chemical classes that are important in differentiating the phenotypes × time, metabolite enrichment analysis was performed on 24 and 48 hpi samples. We found that only 19 MFs were unique to resistant x time, one was unique to the hypersensitive phenotype x time and three were unique to the susceptible phenotype × time interaction. Of these MFs, the identities of only seven could be structurally predicted based on mass spectrometry fragmentation patterns (Table 2). Using PubChem compound summaries, we were able to assign probable compound super/subclasses or pathways to these seven metabolites. Included in these MFs were a range of defence-related classes including fatty acyl glycosides, hormone-related MFs (e.g., sphingolipid/ceramide, phosphatidylglycerol), a compound structurally similar to quorum sensing molecules, as well as a pyrethrin.

Terpenoids Alter with Time Post-Infection but Not with Resistance Phenotype
Time post-inoculation, regardless of resistance phenotype, appears to be the most influential factor with respect to changes within the metabolome. While some of these responses may be associated with leaf aging, as we only used one timepoint for an uninfected control, these results show that A. psidii infection is able to trigger major metabolite responses in M. quinquenervia leaves. We further examined the temporal changes of these significant MFs identified with analysis of variance (ANOVA). Through hierarchical clustering analysis, we found that a majority of these significant MFs were enriched shortly after A. psidii infection (24 or 48 h after inoculation) and gradually reduced at 5 days after inoculation (Figure 2A). Based on these results, we used a k-means clustering approach for Microorganisms 2022, 10, 383 8 of 16 each phenotype to group the metabolites based on the pattern of temporal changes. For each phenotype, we identified two clusters of inducible metabolites at 24 and 48 h post inoculation ( Figure 3A; denoted by red title). Of the 283 MFs identified within these six clusters, a majority are shared by the different phenotypes. The structural classes of these inducible MFs are terpenoids, fatty alcohols, and dipeptides, and included intermediates of phytohormone pathways ( Figure 3B; Supplemental Table S3).

Pre-Infection Enrichment of Flavonoid Compounds Found in Resistant Individuals
To establish if metabolites are important in defining M. quinquenervia resistance to A. psidii before infection, uninfected leaf samples of the same plastochron index sampled at the same time were analysed by principal component analysis; no significant grouping was found according to phenotype ( Figure 4A). Subsequently, PLS-DA and ANOVA analysis was applied to identify if any MFs could significantly separate the three phenotypes ( Figure 4B). The heatmap of the top MFs (ranked by the ANOVA FDR value) shows that resistant individuals were characterised by lower relative abundances of the majority of the MFs identified ( Figure 4C; Supplemental Table S4). To determine the chemical classes of MFs that are associated with resistance to A. psidii before infection, metabolite enrichment analysis was performed. The MFs enriched in the resistant phenotype that were structurally predicted using Progenesis included the chemical classes of flavonoids, cinnamic acids and polyketides ( Figure 4D; Supplemental Table S5).

Discussion
Melaleuca quinquenervia is a suitable species of the Myrtaceae for improving our understanding of the mechanisms related to resistance to A. psidii as individuals within this

Discussion
Melaleuca quinquenervia is a suitable species of the Myrtaceae for improving our understanding of the mechanisms related to resistance to A. psidii as individuals within this species display a range of susceptibility from complete resistance to complete susceptibility.
We sought to use an untargeted metabolomics approach to identify global changes to a broad spectrum of known and unknown metabolites following A. psidii infection and to predict the metabolic basis of infection phenotype. Our findings demonstrate that regardless of rust response phenotype, all individuals tested exhibited temporal changes in the metabolome in response to pathogen challenge. Metabolites found to be associated with pathogen defence in other plant species were found to be more enriched in resistant M. quinquenervia individuals. Furthermore, our analyses suggest that pre-existing differences in metabolic profiles in the leaves of M. quinquenervia prior to infection may serve as an addition means by which we can predict resistance phenotypes.

Melaleuca quinquenervia Synthesises MFs in Response to Pathogen Challenge That Are Independent of Infection Phenotype
Plants are universally responsive to colonisation by microbes, regardless of the degree of damage they may cause [37,38]. Following inoculation, A. psidii urediniospores germinate and form appressoria within the first 18 h [17]. Previous studies in S. luehmannii and E. grandis demonstrated that the host is very responsive transcriptomically at this stage [21,23,24]. In S. luehmannii, the greatest host response was observed 48 h after pathogen challenge, with pathways including a number of secondary metabolic pathways (e.g., flavonoid metabolic enzymes, cell wall active enzymes) [21]. Brassinosteroid signalling was also observed in E. grandis at the same time point [24]. Metabolically, Sekiya et al. (2020) found that the number of differentially regulated MFs of E. grandis during infection stayed consistent, or increased slightly in number, throughout the a range of timepoints through 24 h post inoculation, while in the wheat:Puccinia triticina pathosystem, differential enzyme responses were observed at 48 h post inoculation [39]. Similarly, in the soybean:Phakopsora pachyrhizi interaction, changes in host metabolome increased through 72 h post infection [40]. Our results show that in M. quinquenervia, a metabolomic response was triggered in all plants upon infection, with a large overlap in the MFs regulated between individual plants tested. The metabolomic changes were greatest at 24 hpi, decreasing at 48 hpi and approaching pre-infection levels by day 5. This indicates that a strong chemical defence response to the pathogen is triggered in M. quinquenervia, regardless of phenotype, and that this metabolic response was faster and greater than documented in other model systems considered here [21], while some MF responses may have been resultant of leaf age [25]. Interestingly, a range of terpenoids were among the induced metabolites that did not vary by resistance phenotype. Several studies have considered the role played by terpenoids during host interaction with biotrophic pathogens, however, a clear role of these molecules has yet to be clearly established as some terpenes are associated with disease resistance [33,41,42], while others with susceptibility [30,[43][44][45]. For example, in grapevine challenged with the biotrophic oomycete Plasmopara viticola, a phenotype by time-relationship was found for terpenes in resistant cultivars only where pathogen growth and reproduction were significantly repressed in plant hosts producing higher levels of terpenoids [46]. Our results are, therefore, different from these previous studies and suggest that in M. quinquenervia terpenoids may not be the most significant class of compound associated with defence against A. psidii. Despite this, our data suggest that the host is responsive metabolically to challenge by rust infection, and that small changes to host metabolism following infection must be considered as they may be deterministic in establishing the difference in between what defines a susceptible versus resistant phenotype.

Melaleuca Disease Resistance Involves MFs That Are Both Phenotype-Specific and Responsive to Phenotype × Time
The degree to which a host responds metabolically to colonisation by a pathogenic microbe can determine the disease resistance phenotype. Similarly, an interactive effect between host phenotype and the stage of disease development (i.e., phenotype × time interactive effect; [47]) may play a role in determining the resistance phenotype in a plant. Our study found evidence for metabolites tied to the level of M. quinquenervia resistance to A. psidii that were both phenotype-specific as well as associated in a phenotype × time manner. In the former category, we found that organoheterocyclic compounds, metabolites associated with nucleic acids, organic acids and carbohydrates, were all enriched in resistant individuals regardless of time. The metabolite superclass of organoheterocyclic compounds, including subclasses such as carbazole, porphyrin, and butanolide, are almost universally associated with anti-microbial or antioxidant activities [48][49][50]. Therefore, induction of these compounds by the onset of fungal colonisation within resistant individuals is likely to enhance defence against the pathogen. Compound classes such as carbohydrates, and their putative role in host resistance or susceptibility, are more difficult to interpret. Past work with leaf rust pathogens suggests that carbohydrate rich leaves are more susceptible to infection as they are rich in growth limiting nutrients for the pathogen [51]. This hypothesis, however, is not supported by more recent work that found leaf nutritional status had no impact on the development of Melampsora laricipopulina in the leaves of Populus deltoides × P. nigra, but rather that secondary metabolism differences between leaf ages were the driver of disease outcomes [52]. As we only assayed leaves of consistent age within our analyses, it is likely that differences in carbohydrates are tied to phenotypic variation between individuals and future work will be important to understand the roles of these metabolites in host resistance.
Of the metabolites that exhibited a link to disease resistance in a phenotype by time interaction, structural-based MF prediction identified a number of metabolites putatively associated with phytohormone defence pathways, with two MFs linked to salicylic acid biosynthesis and signalling (Table 2; [53,54]. SA has been shown to have a primary role in resistance to rust pathogens including M. larici-populina:Populus [55], Puccinia triticina [39] and Uromyces spp. [56]. Recent work suggesting that SA is involved in localised repression of haustorial formation [57] would explain why we do not observe physical damage within resistant lines of M. quinquenervia. A quorum sensing molecule, methyl 3-hydroxypalmitate, was also identified as increasing in concentration with time only in resistant individuals. Quorum sensing compounds have previously been characterised as being pathogen-produced (e.g., Rolstonia solanacearum; [54]) and necessary for the progression of disease. The discovery of this MF was unanticipated because it was induced in resistant M. quinquenervia individuals without visible disease symptoms. It is possible that the compound was host-produced as other research suggests that quorum sensing-like molecules can be produced by plants to recruit biocontrol bacteria [58]. Therefore, in addition to considering the specific MFs produced in plants during disease challenge, the timing of their production must also be considered as different MFs will have roles in host immunity by affecting development of the pathogen from spore germination through to reproduction.

Basal Metabolic Processes May Contribute to Myrtle Rust Resistance
In a range of pathosystems, research has sought to predict disease resistance ratings of plant varieties based on basal metabolism in the absence of the disease-causing microbe. These research strategies are based on the hypothesis that steady states of different physical or chemical defence compounds may be naturally higher in some individuals, thereby affording them greater protection against the initial stages of disease. Physical barriers associated with cuticular waxes enriched in hexadecenoic acid relate positively to increased rust resistance in E. grandis and Eucalyptus phaeotricha [32], while elevated levels of limonene have also been associated with rust resistance [33]. We also detected consistent differences in the metabolomic profile of the individual M. quinquenervia plants prior to infection that significantly predicted disease resistance ratings. In particular, flavonoids were enriched in the resistant plants. A number of studies have highlighted that flavonoids may inhibit disease progression by inhibiting germination of fungal spores or fungal growth [59,60], while others have shown an opposing role [61,62]. Other plant studies have shown that flavonoids are important in the resistance of plants to biotic stress, for example potato leaves resistant to late blight were enriched with higher concentrations of some flavonoids [63]. In resistant E. grandis flavonoids are increased in abundance in resistant lines 24 hpi, which was further supported by increased abundance of phenylpropanoid and flavonoid biosynthetic proteins [25]. Similarly, transcriptomic analysis in Syzygium luehmannii identified elevated expression levels of the enzyme phenylalanine ammonia lyase in resistant plants, which supplies the precursor for flavonoid metabolism [21]. Therefore, while the exact flavonoids found in our work are predictions only, flavonoid molecules present within certain M. quinquenervia individuals appear to be major determinants in myrtle rust resistance.

Conclusions
Altogether, our research suggests that the resistance to A. psidii in M. quinquenervia is mediated in part by two separate pools of metabolites: those that are basally produced in leaves and those that are pathogen inducible. The similarity in metabolite classes found to predict myrtle rust resistance in M. quinquenervia to those found to predict resistance to biotrophic pathogens in a broad range of plants provide promise that these MFs could be resistance biomarkers that are not just relevant within Melaleuca, but also to other Myrtaceae. Further studies are required, therefore, to determine if this is the case and, where a consistent subset of biomarkers is identified, adapted to enable screening for resistant germplasm in nursery breeding programs to protect the biodiversity in this ecologically, culturally, and economically important plant family.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/microorganisms10020383/s1, Table S1: The top 500 molecular features that were significantly changed in abundance during the experiment, Table S2: Enriched molecular features that differentiated the phenotypes during early stages of infection with Austropuccinia psidii, Table S3: Structural prediction of molecular features enriched in Melaleuca quinquenervia leaves during the early stages of infection by Austropuccinia psidii, Table S5 Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: Links to publicly available data will be provided upon acceptance of the manuscript for publication.

Acknowledgments:
The authors would like to acknowledge the Mass Spectrometry Facility (MSF) of Western Sydney University for access to its instrumentation and staff. The authors are grateful for the assistance of Meena Mikhael at the Mass Spectrometry Facility.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.