Effect of cultivation media and temperature on metabolite profiles of three nematicidal Bacillus species


Globally, root-knot nematode (RKN) infestations cause great financial losses. Although agrochemicals are used to manage these pests, there is increased interest in using biocontrol agents based on natural antagonistic microorganisms, such as Bacillus. These nematicidal bacteria demonstrate antagonism towards RKN through different modes of action, including specialised metabolite production. The aim of this study was to compare metabolite profiles of nematicidal Bacillus species and assess the influence of cultivation conditions on these profiles. Two hyphenated metabolomics platforms, gas chromatography-mass spectrometry (GC-MS) and liquid chromatography coupled to quadrupole time-of-flight mass spectrometry (LC-QTOF-MS), were employed to profile and compare metabolite features produced during the cultivation of three nematicidal Bacillus species (Bacillus firmus, B. cereus and B. soli) in complex Luria-Bertani broth (LB) and a simpler minimal broth (MB), at three different temperatures (25, 30 and 37°C). Cultivation in complex LB as opposed to simpler MB resulted in the production of more statistically significant metabolite features. Selected temperatures in this study did not have a significant influence on metabolite profiles. Moreover, media-specific influences outweighed temperature-specific influences on metabolite profiles. Results from this study are a valuable first step in establishing suitable cultivation conditions for the production of Bacillus metabolites of interest.

Root-knot nematodes (RKN; Meloidogyne species) are microscopic soil-borne phytopathogens that severely damage the roots of agricultural crops (Cao et al., 2019) by interfering with water and nutrient uptake from surrounding soil (Gao et al., 2016). This leads to the formation of giant, multi-nucleate root cells (galls) that serve as specialised feeding sites for RKN (Abbasi et al., 2014;Phani et al., 2021). Nematode infestations reduce overall crop quality and can cause 100% crop loss (Phani et al., 2021) amounting to $US 125 billion annually (Mesa-Valle et al., 2020). Although chemical nematicides have been effectively used to manage these infestations (Rosas-García, 2009;Ebone et al., 2019), there are concerns about the negative impact of chemicals on human and environmental health . Natural antagonistic organisms or nematicidal compounds derived from these organisms are a possible alternative to manage phytopathogens (Cawoy et al., 2011;Timper, 2014;Gao et al., 2016;Dadrasnia et al., 2020;Mazzuchelli et al., 2020).
Bacillus-nematode antagonism has been attributed to Bacillus secondary or 'specialised' metabolite production (Xiong et al., 2015;Gao et al., 2016;Tobias & Bode, 2019;Sharrar et al., 2020). However, the identities and characteristics of many of these metabolites remain largely unknown, with studies often failing to elaborate on how the specialised metabolites were distinguished from other metabolites (Horak et al., 2019). This is concerning since bacterial metabolism not only involves metabolite formation, but also conversion of metabolites into different compounds, underlying the importance of cultivation conditions. Culture media (complex media vs simple media), cultivation method (batch culture vs continuous culture), temperature, salinity, initial media pH and oxygen availability influence bacterial metabolite and hydrolytic enzyme production (Horak et al., 2019). Although Bacillus species produce a plethora of specialised metabolites (Akeed et al., 2020), the production of these specialised metabolites is not always conserved since different gene clusters can be present within different species or strains (Palazzini et al., 2016).
While Bacillus species harbour multiple biosynthetic gene clusters (BGCs) (Tobias & Bode, 2019) for specialised metabolite production in their genomes, it is well known that many cryptic BGCs are often not expressed under typical laboratory conditions (Loureiro et al., 2018;Singh et al., 2019). Therefore, specialised metabolite production is highly dependent on not only environmental factors (pH, moisture, temperature; Tyc et al., 2017) but also cultivation conditions (Sidorova et al., 2020). It is thus important to note that certain environmental cues (e.g., specific culture medium components and cultivation temperature) can lead to the expression of silent BGCs (Loureiro et al., 2018). Currently, investigation into optimal cultivation conditions for strains producing bioactive molecules is an important avenue of research (Sidorova et al., 2020).
Metabolomics provides a snapshot of the metabolites present within a biological system at a certain time and under specific conditions (Oyedeji et al., 2021). It can help bridge the gap between a microorganism's genotype (inherited genetic material), phenotype (observ-able characteristics) and chemotype (differences in specialised metabolites within the same species/subspecies) (Polatoglu, 2013;Orgogozo et al., 2015;Grim et al., 2019). Metabolomics has also been used to identify plant metabolites associated with RKN resistance in watermelons (Kantor et al., 2018) and tomatoes (Afifah et al., 2019), highlighting its potential in the field of nematology. If suitable cultivation conditions are developed for known nematicidal Bacillus species, metabolite features or inhibitory enzymes could be identified through metabolomics. However, the development and registration of biocontrol agents are challenging (Valivelli et al., 2014), especially formulations that contain 'whole organisms'. Certain characteristics contained within whole organisms are not favourable for use in commercial biocontrol formulations, including toxins produced by B. cereus, which have been implicated in food poisoning (Granum & Lund, 1997). As a result, the end goal of identifying inhibitory metabolites/enzymes would not be to use whole organisms as an inoculum in a biocontrol agent, but rather either: i) to extract the metabolite/enzyme of interest for biocontrol use (Gao et al., 2016); or ii) to synthesise the metabolite/feature in a laboratory using biotechnology. A well-known example of the successful application of biotechnology in pesticide research is the large-scale production of biopesticides containing B. thuringiensis Cry proteins exhibiting insecticidal activity against different agricultural pests (Rosas-García, 2009). Bacillus-based biocontrol agents (BCA) used to manage nematode infestations (Engelbrecht et al., 2018) will improve the ecological integrity of agricultural ecosystems.
Since a previous review article by the authors highlighted the importance of cultivation conditions in bionematicide research (Horak et al., 2019), the objectives of this work were to employ an untargeted metabolomics approach, using gas chromatography-mass spectrometry (GC-MS) and liquid chromatography coupled to quadrupole time-of-flight mass spectrometry (LC-QTOF-MS), to investigate the effect of media composition (complex media vs simple media) and cultivation temperature (25, 30 and 37°C) on the metabolite profiles of three agriculturally important Bacillus species: B. firmus, B. cereus and B. soli. Differences between the metabolite features produced by nematicidal Bacillus species compared to a nonnematicidal control bacterium, Escherichia coli OP50, were also evaluated.

MEASUREMENTS
After incubation, all species were harvested when they were 20% into their stationary phase. This was followed by measuring the optical density (OD) of all cultures in triplicate at 600 nm (OD 600 ) (Wu & Li, 2016), using a carousel Jenway 7300 spectrophotometer (Lasec). Noninoculated broth was used as a blank. The OD was measured for data normalisation (Worley & Powers, 2013;Venter, 2018).

MEDIUM
To obtain the cell-free supernatant of the Bacillus species and thus investigate the extracellular metabolite features, the intact bacterial cells were separated from the culture medium by centrifugation at a low temperature (Mashego et al., 2007). This was followed by sterile filtration of the media using 5 ml syringes and 0.45 μm syringe filters, followed by 0.22 μm syringe filters. For metabolomics analysis, 100 μl aliquots of each filtrate were prepared on ice and biological replicates of six (n = 6) were included. All samples were capped and stored at −80°C to retain the original metabolite properties until further processing and metabolomics analysis (Pinu & Villas-Boas, 2017).

PREPARATION OF QUALITY CONTROL SAMPLES
To ensure no observed batch effect (Venter, 2018), pooled quality control (QC) samples were included. These samples were representative of the qualitative and quantitative composition of all samples prepared for both GC-MS and LC-QTOF-MS analyses and consisted of small aliquots of each of the bacterial filtrates (B. firmus, B. cereus, B. soli and E. coli OP50) (Dudzik et al., 2018).

ANALYSIS
Before GC-MS analysis, all samples were lyophilised to deactivate enzymatic activity and concentrate the extracellular metabolites (Pinu & Villas-Boas, 2017), after which samples underwent a two-step derivatisation approach of methoxyamination, followed by silylation. The lyophilised samples were derivatised by the addition of 50 μl methoxyamination solution (20 mg ml −1 methoxyamine hydrochloride dissolved in anhydrous pyridine), vortexed for 30 s and then incubated at 60°C for 1 h. After incubation and subsequent cooling, 50 μl of N,O-bis(trimethylsilyl) trifluoroacetamide containing 1% trimethylchlorosilane (BSTFA-TMCS; 99:1) (Merck Life Sciences) was added to the sample before vortexing and then incubating at 60°C for 1 h. After incubation, the samples were allowed to reach room temperature. Subsequently, 50 μl of iso-octane containing the FAME (fatty acid methyl esters) mix, according to the method of Fiehn (2017), was added to samples and samples were vortexed for 30 s. The solution was transferred with a Pasteur pipette to a new 250 μl pulled point glass insert, fitted in the screw-top vial and capped. The samples were then submitted to the laboratory for GC-MS analysis. For each batch of GC-MS samples, both blank and QC samples were prepared as described above and included in the analysis.
For LC-QTOF-MS analysis, samples were prepared by the addition of 300 μl of acetonitrile to the microcen-trifuge tubes containing the bacterial filtrates. This mixture was left to thaw, after which it was vortexed briefly and then centrifuged at 4°C and 13 500 g for 15 min. A 100 μl volume of the acetonitrile/filtrate mixture was transferred to a screw-top vial fitted with a 400 μl flat bottom glass insert, already containing 100 μl internal standard mix. The vials were capped and then submitted to the laboratory for LC-QTOF-MS analysis. For each batch of LC-QTOF-MS samples, both blank and QC samples were prepared as described above and included in the analysis.

METABOLITE FEATURES
An Agilent GC-MSD (Agilent Technologies) consisting of a 7890A gas chromatograph with a split/splitless injector (250°C), equipped with a 7683B autosampler coupled to a 5975B inert XL mass selective detector was used. The gas chromatograph was equipped with a 30 m Agilent DB-5MS column, 0.25 mm i.d. and 0.25 μm film. For the carrier gas helium was used at a flow rate of 1.0 ml min −1 . A sample volume of 1 μl was injected with a 10:1 split ratio. The initial oven temperature was 60°C for 1 min, ramping to 325°C at 10°C min −1 followed by a hold time of 10 min, resulting in a total runtime of 37.5 min per sample. The transfer line temperature was set to 290°C while the source temperature was set at 230°C. The acquisition was delayed for the first 5.9 min serving as a solvent delay. Data were captured with an acquisition rate of 20 spectra (50-600 m/z) s −1 , with EI energy of 70 eV (Fiehn, 2017).
The Agilent 1200 LC-QTOF-MS system with a Cogent diamond hydride 2.1 mm × 10 mm, 4 μm particle size column (MicroSolve Technologies) was used and the temperature maintained at 50°C. The mobile phases consisted of (A) water containing 0.1% acetic acid (v/v) (pH 3.4) and 0.1% ammonium acetate (w/v) and (B) 90% acetonitrile (v/v) with 0.1% ammonium acetate (w/v) and 0.1% acetic acid (v/v). For optimum chromatographic separation, the following gradient was used: starting at 100% B, then linearly decreasing to 40% B over 10 min, followed by a 1 min hold at 40%. The gradient then returned to 100% B and the column was re-equilibrated at 100% B for 6 min. Mass spectrometry was performed using an Agilent 6510 QTOF (Agilent) with an ESI source in positive ESI. A drying gas temperature of 300°C with a gas flow of 10 l min −1 and nebuliser pressure of 310 kPa, a capillary voltage of 4000 V, fragmentor voltage of 150 V and a skimmer voltage of 65 V were used. The instrument was operated in the extended dynamic range mode and data collection in the m/z range of 70-1700 amu. For accurate mass reference a reference solution containing purine (mass 121.050873 [M + H] + ) and hexakis (1H, 1H, 3Htetrafluoropropoxy) phosphazine (mass 922.009798 [M + H] + ) was constantly infused (Callahan et al., 2009).

STATISTICAL ANALYSIS
Statistical analysis of the metabolomics data involved the following: i) data extraction; ii) data pre-processing; iii) normalisation; and iv) data pre-treatment (Venter, 2018). Data extraction of LC-QTOF-MS data was performed using Agilent's MassHunter Qualitative software (Version B.06). The find by molecular feature (FbMF) extraction algorithm was used according to Agilent's specifications to aid with data extraction. Agilent's Mass Profiler Professional (MPP) (Version B.02.02) was used to align the data. GC-MSD spectra were identified by spectral analysis, using the Automated Mass Spectral Deconvolution and Identification System (AMDIS, Version 2.7) and the NIST 2011 mass spectral library. A match factor of 80% was set. Data files created in AMDIS were imported into Agilent's MassHunter MPP (Version B.02.02), where chromatographic retention time aligned were performed across multiple data (Venter et al. 2015;Willers et al., 2016). Data extraction was followed by preprocessing during which instrument-specific data were edited and relevant analytical information extracted. Zerofiltering was used in which defined variables without any statistical and/or biological value were removed through the elimination of features with a vast amount of missing values (Venter et al., 2015); i.e., only features present in 100% of one case were included for further statistical analysis, with one case in this study referring to a single Bacillus species (all six replicates of a condition).
Biomass normalisation was performed using the OD 600 values of the bacterial cultures measured directly after incubation to remove systematic variation of a nonbiological origin among samples within a dataset (Kuligowski et al., 2014). The resulting data matrix was analysed using MetaboAnalyst 5.0. (https://www. metaboanalyst.ca) Xia & Wishart, 2016;Chong et al., 2019). Data integrity check and missing value estimation were performed to ensure that the data met the basic requirements for meaningful downstream statistical analysis (Chong et al., 2019). During data processing, the data were transformed to ensure that: i) the variables exhibited a more normal or Gaussian distribu-tion; and ii) were in a usable format for statistical analysis. In this study, log transformation was applied followed by autoscaling to ensure that the variables are comparable to one other (Chong et al., 2019). To identify statistically significant (P < 0.05) variables among experimental conditions, univariate analyses included Student's t-test and one-way analysis of variance (ANOVA), followed by Fisher's least significant difference (LSD) post-hoc analysis, while multivariate statistical analysis (MVA) included principal component analysis (PCA).

COMPARISON OF Bacillus SPECIES TO Escherichia coli OP50
Cell-free filtrates of three nematicidal Bacillus species (B. firmus, B. soli and B. cereus) and a negative control bacterium, E. coli OP50, were subjected to GC-MS and LC-QTOF-MS analyses. Results obtained from LC-QTOF-MS analysis showed similar, but less distinctive, grouping of Bacillus metabolite features, compared to GC-MS analysis. Consequently, only GC-MS results are shown.
Univariate and multivariate analyses were performed on the resulting GC-MS data matrix for cultivation in both LB and MB after data filtering, normalisation and pre-processing had been completed. The scores plot (with semi-transparent confidence intervals of 95%) obtained from untargeted GC-MS analysis, indicate the distribution of metabolite features of Bacillus species compared to control bacterium, E. coli OP50, when cultivated in LB and MB at 25, 30 and 37°C over 7 days and sampled according to their stationary phase (results not shown In this study, two types of cultivation media were investigated: a complex LB and a simpler MB. Figure 1 illustrates the 2-D PCA scores plots obtained from the untargeted GC-MS analysis of B. firmus, B. cereus and B. soli when cultivated in LB and MB (25, 30 and 37°C) over 7 days, and sampled during the stationary growth phase. Quality control samples are also included and clustered together on the scores plot in both Figure 1A and 1B. Three distinct groups were observed for cultivation in LB (Fig. 1A): i) B. firmus and B. cereus; ii) B. soli; and iii) QC samples, with a cumulative variance of 60.1% for the first two PC axes. Metabolite features of B. firmus and B. cereus grouped closer together (with subtle variation in their metabolite profiles) than to that of B. soli. By contrast, cultivation in MB (Fig. 1B) did not result in any distinct clustering of Bacillus species, with PC1 and PC2 indicating a variance of 72.5% and 7.4%, respectively. High sample-to-sample variability was observed for B. soli (Fig. 1A), and B. firmus and B. cereus (Fig. 1B). Overall, cultivation in LB resulted in more distinct metabolite clustering than cultivation in MB. Figure 2 shows the one-way ANOVA (followed by Fisher's LSD post-hoc analysis) obtained from the untargeted GC-MS analysis of bacterial cultures cultivated in LB and MB at 25, 30 and 37°C over 7 days. Samples were collected during the stationary growth phase. Red plots are the statistically significant (P < 0.05) metabolite features detected, whilst green plots indicate non-statistically significant (P > 0.05) features. More statistically significant (P < 0.05) features were produced during cultivation in LB (264) ( Fig. 2A), compared to MB (63) (Fig. 2B). Figures 3 and 4 show the scores plots of Bacillus species cultivated in LB and MB, respectively, at 25, 30 and 37°C (over 7 days). Samples were collected during the stationary growth phase. Tighter clustering and a smaller semi-transparent confidence interval were observed for B. firmus when cultivated in LB at 30°C (Fig. 3A), compared to cultivation at 25 and 37°C. There are similarities in the metabolite profiles for cultivation at 25 and 30°C since semi-transparent confidence intervals overlap (Fig. 3A). Cultivation at 37°C showed slightly different metabolite clustering. The first two PC axes   indicated a cumulative variance of 51.3%. No distinct clustering patterns were observed for B. firmus in MB (Fig. 4A); PC1 and PC2 accounted for 63.6% and 12.8% of the variance, respectively. No differentiation was observed in the profiles of B. soli at 25, 30 or 37°C in LB (Fig. 3B), with 58.6% of the variance indicated along the first PC and 11.1% on the second PC. Metabolites produced by B. soli in MB at 25 and 30°C were more similar than those produced at 37°C, due to a closer association on the scores plot (Fig. 4B). Cultivation at 25 and 30°C showed low sample-to-sample variability, compared to high sample-to-sample variability of cultivation at 37°C. The first two PC axes accounted for a cumulative variance of 83.3%.

THE ROLE OF CULTIVATION TEMPERATURE ON THE PRODUCTION OF Bacillus METABOLIC FEATURES
There was a closer association between B. cereus metabolites produced in LB cultivated at 30°C (condensed clustering) and 37°C (Fig. 3C), with overlapping semiconfidence intervals. A cumulative variance of 51.7% is reported. Similarly, to B. firmus cultivated in MB (Fig. 4A), no clear groupings were observed for B. cereus metabolites cultivated in MB (Fig. 4C). Along PC1 and PC2, 73.7% and 8.0% variance were accounted for, respectively.
Important metabolite features (mz and rt) identified by one-way ANOVA (followed by Fisher's LSD posthoc analysis) of Bacillus species cultivated at 25, 30 and 37°C in LB and MB over 7 days are shown in Figures 5  and 6, respectively. Cultivation in LB at 25°C (Fig. 5A) and 30°C (Fig. 5B) yielded 254 and 242 statistically  soli and B. cereus when cultivated in MB at three different temperatures over 7 days and sampled according to their stationary phase, including the QC samples. The P value threshold is 0.05. ANOVA was followed by post-hoc analyses based on Fisher's LSD. Red plots indicate statistically significant features (P < 0.05), while green plots indicate non-significant features (P > 0.05); A: Features identified for cultivation at 25°C; B: Features identified for cultivation at 30°C; C: Features identified for cultivation at 37°C. significant (P < 0.05) metabolite features, respectively. Fewer features (164) (P < 0.05) were produced when Bacillus species were cultivated in LB at 37°C (Fig. 5C). A similar trend was observed for cultivation in MB; more statistically significant (P < 0.05) metabolite features (87) were produced for cultivation at 25°C (Fig. 6A) and 30°C (Fig. 6B), than at 37°C (63) (Fig. 6C).

METABOLITE FEATURES PRODUCED BY Bacillus SPECIES AND Escherichia coli OP50
The key to understanding the biological function and possible biotechnological applications of metabolites are well-characterised metabolomes . Metabolomics provides a means of understanding the in vitro metabolic activities of microorganisms and how they relate to changes in cultivation media and temperature. It plays a key role in novel microbial compound discovery (Nguyen et al., 2012). Features are spectrometric signals (retention time or intensity) belonging to metabolites that have not yet been identified (Van Dam & Bouwmeester, 2016). They are fundamentally related to concentrations of a specific metabolite (Alonso et al., 2015). Since metabolomics allows microorganisms to be clustered together based on the types/concentrations of metabolites they produce (Breitling et al., 2013), statistically significant differences between features of nemati-cidal bacterial strains and a control bacterium might indicate the cause of the observed activity.
In this study, the cell-free filtrates of three nematicidal Bacillus species, B. firmus, B. soli and B. cereus, and a negative control bacterium, E. coli OP50, were subjected to GC-MS and LC-QTOF-MS analyses. Bacillus firmus and B. cereus were selected for their known nematicidal activity (Gao et al., 2016;Geng et al., 2016;Ghahremani et al., 2020), while B. soli is novel in terms of nematode virulence. Although described and validly published 15 years ago after isolation from disused hay fields in The Netherlands (Heyrman et al., 2004), B. soli has only recently been evaluated for its nematicidal potential against Meloidogyne incognita second-stage juveniles (Engelbrecht et al., 2020). The negative control bacterium, E. coli OP50, was selected as it is the preferred bacterium fed to nematodes under laboratory conditions (Stuhr & Curran, 2020).
Based on their clustering patterns, the three Bacillus species produced different metabolite features or different concentrations of the same features than E. coli OP50 (results not shown). According to Worley & Powers (2013), there is inherent variability in each sample during metabolomics analysis, even if samples belong to the same species. It is therefore important to determine whether sample-to-sample variability can be attributed to inherent variability or simple methodological differences during sample preparation. The closer association between the metabolite features of B. firmus and B. cereus with E. coli OP50, when cultivated in MB, could likely be attributed to the high sample-to-sample variability. Due to the clear differentiation between the metabolite profiles of Bacillus species and E. coli OP50 cultivated in LB and MB (25, 30 and 37°C over 7 days), E. coli OP50 was excluded from any further results.

THE INFLUENCE OF CULTIVATION MEDIA ON Bacillus METABOLITE PRODUCTION
Results demonstrated that the cultivation medium influences the type and/or concentration of metabolite features produced by the three studied Bacillus species. The tight clustering of QC samples for cultivation in both LB and MB indicates no observed batch effect and accuracy of the metabolomics analysis (Dudzik et al., 2018). When cultivated in LB, features produced by B. firmus and B. cereus were more similar or had similar concentrations. Bacillus soli demonstrated differences when compared to B. firmus and B. cereus. Comparable results were obtained by Perez-Fons et al. (2014), with five species more closely associated (B. subtilis PY79, B. subtilis 'natto ', B. coagulans-BC30, B. licheniformis-Biosporin 2336, B. subtilis-Biosporin 2335, whilst the sixth species grouped separately (B. indicus HU36). In this study, the less distinctive grouping of metabolite profiles of Bacillus species cultivated in MB is attributed to the simpler composition of MB, presence of outliers and different cultivation temperatures of the samples.
Culture parameters (cultivation media, temperature, pH and incubation period) influence metabolite production due to differences in nutrient content and resulting metabolic pathways followed (Horak et al., 2019). Sidorova et al. (2020) also emphasised that the degree of bioactive metabolite accumulation not only depends on the producing strain but also significantly on medium composition and pH, as well as the incubation temperature and period. Therefore, the time of sampling is also important -in this study, sampling during the stationary growth phase was suitable for the detection of metabolite features. Favre et al. (2017) investigated the influence of culture parameters on cellular metabolism and global inter-strain metabolic discrimination of four marine bacteria (Persicivirga mediterranea TC4 and TC7, P. lipolytica TC8 and a Shewanella sp.). Results showed the importance of growth phase, since more than two-thirds of the m/z signals were detected in higher amounts during the stationary phase. Greater chemical metabolite diversity was also observed during this phase (Favre et al., 2017). In theory, heterologous metabolite production would be maximised during stationary phase growth, as opposed to exponential phase growth during which a trade-off between biomass and product synthesis would occur (Chubukov & Sauer, 2014). This highlights the importance of stationary phase growth for optimum metabolite production, especially in bionematicide research.
The chosen cultivation medium can strongly influence specialised metabolite production and sampling efficiency (Frisvad, 2012;Meyer et al., 2013). The latter is an important aspect of microbial metabolomics since sample preparation can affect the quality of metabolic profiles, coverage of metabolites, the accuracy of final metabolomics results and conclusions made (Raterink et al., 2014;Xu et al., 2014;Patejko et al., 2017). When fewer nutrients are available for utilisation (e.g., MB), the metabolism is limited due to less metabolic activity. Consequently, bacteria could produce more similar metabolites due to nutrient limitations. Complex media (e.g., LB), contain components of natural origin such as plant and/or animal extracts (peptone, tryptone, meat extracts, casein hydrosalyte) (Diederichs et al., 2014), together with sugars (e.g., glucose) as energy sources (Ratiu et al., 2017). Additionally, yeast extract is one of the most commonly used media constituents and is also present in LB (Diederichs et al., 2014). It is composed of an undefined mixture of amino acids, carbohydrates, peptides and other trace elements (Diederichs et al., 2014;Sørensen & Sondergaard, 2014). Bacteria, therefore, have an overabundance of nutrients to utilise until depletion, resulting in a reduction of bacterial cells. When bacterial cultures start to utilise a new resource, cells density and metabolite production will increase. By contrast, minimal broth only contains the most essential nutrients for survival (Hall et al., 2013).
The results of this study are in agreement with several other metabolomics-based studies Ratiu et al., 2017;Lim et al., 2018).  found that cultivation of E. coli MG1655 in complex medium (Luria broth) caused higher growth rates and final cell densities compared to cultivation in minimal medium (M9). Escherichia coli (ATCC8739) only produced indole when cultivated in a more complex trypticase soy broth (TSB), as opposed to cultivation in Mueller-Hinton (MH) or a minimal salts (M9) medium (Ratiu et al., 2017). Additionally, media-dependent production of the specialised metabolite, prodignine by Streptomyces coelicolor A3(2) was observed by Lim et al. (2018).
Different carbon and nitrogen sources in cultivation media also affect the resulting extracellular metabolite profiles obtained. Abd Rahim et al. (2019) found that the presence of a complex nitrogen source (yeast extract) resulted in the highest yield of Aspergillus terreus metabolites, as opposed to a medium containing simpler nitrogen sources. When investigating the metabolome of B. subtilis grown in the presence of different carbon sources (including malate, fumarate and citrate), results revealed different extracellular metabolite profiles based on the availability of different carbon sources (Meyer et al., 2014).
Results obtained for cultivation in LB could be a reflection of the phylogenetic and evolutionary relatedness of Bacillus species. Based on 16S rRNA gene sequencing, B. firmus and B. cereus are more evolutionarily related (Logan & De Vos, 2015). Specialised metabolites of bacterial species generally show correlation with taxonomical similarity (Lu et al., 2014). However, this is not always the case as some studies have found metabolite differences amongst species of the same genus (Drapal et al., 2014;Chun et al., 2019). Although the results of this study show a close association between the features of B. firmus and B. cereus during cultivation in LB, different nematicidal metabolites have been isolated and identified from these species (Oliveira et al., 2014;Gao et al., 2016;Geng et al., 2016). This indicates that although similar features can be produced by different species, they do not necessarily exhibit the same bioactivity or are produced in different quantities within species.
In this study, low-level metabolite identification (Level 5) was possible (Schymanski et al., 2014), since the exact mass (m/z) and retention time of features were obtained. This is an important precursor step for higherlevel identification (molecular formula and confirmed structure) (Schymanski et al., 2014), and relates to the main goal of metabolomics which is to identify a few unique chemical features that define a biological system (Worley & Powers, 2013). Higher-level identification has been achieved for some Bacillus nematicidal metabolites, including styrene produced by B. mycoides (Luo et al., 2018), 2-methylbutyric acid by B. pumilus L1 (Lee et al., 2016), uracil by B. cereus and B. subtilis, and dihydrouracil and 9H-purine by B. subtilis (Oliveira et al., 2014).

THE ROLE OF CULTIVATION TEMPERATURE IN THE PRODUCTION OF Bacillus METABOLIC FEATURES
Incubation temperature can also influence metabolite production. Since more condensed clustering was observed for B. firmus when cultivated in LB at 30°C, it suggests that cultivation at this temperature was more repeatable compared to cultivation at 25 and 37°C. No clear groupings were observed for the metabolite profiles of B. firmus when cultivated in MB at the three temperatures, indicating that incubation temperature did not play a major role in metabolite production in this simpler media. This was also the case for cultivation of B. soli in LB and B. cereus in MB. Although temperature influences have been found to play a role in hydrolytic enzyme (Akeed et al., 2020) and specialised metabolite production (Bizani & Brandelli, 2004), the temperatures selected in this study did not have a significant influence on metabolite production of nematicidal Bacillus species. Overall, the metabolite profiles of B. firmus, B. soli and B. cereus were similar when cultivated in LB and MB at 25, 30 or 37°C due to their less distinctive clustering and large, overlapping semi-transparent confidence intervals. More statistically significant (P < 0.05) metabolite features were produced in LB and MB at 25 and 30°C, compared to cultivation at 37°C. However, when Akeed et al. (2020) investigated the influence of incubation temperature on enzyme production of B. licheniformis B307, results showed that temperature significantly influenced production. The maximum chitinase yield was obtained for cultivation at 30°C. Apart from the effect of cultivation temperature on the type/concentration of metabolite features, it can also influence the bioactivity of bacterial metabolites and hydrolytic enzymes of interest. This was evidenced by Lee et al. (2014) when investigating the antagonistic activity of chitinase produced by Lysobacter capsici YS1215, against M. incognita. The highest enzyme activity was exhibited at 40°C.

TIME-OF-FLIGHT MASS SPECTROMETRY ANALYSIS
The use of GC-MS and LC-QTOF-MS platforms was combined in this study as they are considered complementary approaches in metabolomics studies and can provide better coverage of metabolomes of interest (Fiehn, 2017). However, the less distinctive grouping of Bacillus metabolite profiles obtained from LC-QTOF-MS analysis could be attributed to simple sample preparation and clean-up steps in the current protocol before analysis. Gas chromatography-mass spectrometry requires sample derivatisation and therefore offers a higher resolution. Moreover, chromatographic separation and ionisation for LC-QTOF-MS analysis were problematic due to the media matrixes (LB and MB) under investigation. The high content of chemically undefined media constituents present in complex LB and high salt content of simpler MB could have interfered with the chromatographic separation and ionisation during LC-QTOF-MS analysis. For the approach followed in this study, GC-MS analysis is preferred as it yielded more distinctive and reproducible results (Qiu & Reed, 2014;Hu et al., 2018). As a result, LC-QTOF-MS analysis was not pursued any further. However, for any future investigations, the use of LC-QTOF-MS can be optimised by additional sample preparation and clean-up steps, such as soft derivatisation to enhance mass spectrometry signals (Halket et al., 2005).
In conclusion, untargeted metabolomics was employed to investigate the metabolite features produced by three nematicidal Bacillus species cultivated under a specific set of cultivation conditions. This was done to determine whether cultivation media and temperature influenced the metabolite profiles of B. firmus, B. cereus and B. soli. From the results obtained in this study, it is evident that there are both similarities and differences between the metabolite profiles of these species cultivated in LB, indicating that the metabolites potentially responsible for the known nematicidal activity are not necessarily the same metabolite. Moreover, since this study only focused on low-level identification, it remains unclear whether the metabolites detected are responsible for the observed activity. Cultivation in LB yielded more statistically significant (P < 0.05) features, as opposed to MB. This suggests that the production of metabolites could be dependent on the chosen culture medium, and media-specific metabolite production possibly outweighed species-specific production. Distinctive clustering patterns were not observed for Bacillus species when incubated at the three different temperatures.
In addition to temperature influencing metabolite and hydrolytic enzyme production, many classes of Bacillus specialised metabolites, enzymes and toxins have not yet been evaluated for their nematicidal activity (Susič et al., 2020). As a result, both these avenues of interest warrant further research. Previous work has shown that B. soli cell-free filtrates cultivated in LB at 30°C caused statistically significant mortality of M. incognita secondstage juveniles (Engelbrecht et al., 2020). As a next step, it is recommended that in vitro bioassays with nematodes be performed using B. firmus, B. soli and B. cereus cell-free filtrates obtained from cultivation conditions employed in this study to investigate whether the species exhibit bioactivity towards RKN under these conditions. If bioactivity is observed and unique features responsible for the nematicidal activity are produced, higher-level (molecular formula and confirmed structure) identification of these features should be pursued using targeted metabolomics. Overall, this study was a valuable first step in establishing suitable cultivation conditions for the production of Bacillus metabolites of interest. The approaches followed in this study could be employed in other avenues of biocontrol-related research.