1 Executive Summary

This report presents untargeted LC-MS metabolomics analysis of Bacillus amyloliquefaciens BNC5 pellet extracts comparing three conditions: BNC5 alone, BNC5 co-cultured with Ralstonia solanacearum whole cells (BNC5RSw), and RSw alone (dead cell background control).

Key findings:

  • 7744 features showed significant variation across groups (ANOVA, BH-adjusted p < 0.05)
  • 2847 features were elevated in BNC5RSw vs BNC5 (log2FC > 0.58, p_BH < 0.05)
  • After removing 1314 RSw background features, 1533 true BNC5 response features were identified
  • 857 features showed greater than 10-fold elevation in BNC5RSw vs BNC5
  • 1559 features were significantly reduced in BNC5RSw compared to BNC5 alone

2 Study Design

Table 1: Experimental design and sample overview
Condition Description Replicates Role
BNC5 B. amyloliquefaciens BNC5 alone — negative control 6 (3 x Lot1 + 3 x Lot2) Baseline BNC5 metabolome
BNC5RSw BNC5 co-cultured with R. solanacearum whole cells — treatment 6 (3 x Lot1 + 3 x Lot2) RSw-induced metabolic response
RSw R. solanacearum whole cells alone — background control 6 (3 x Lot1 + 3 x Lot2) Background subtraction control

Instrument: HPLC-QTOF ESI-MS (positive ionisation mode) Sample type: Bacterial cell pellets Biological lots: 2 independent preparations per condition Total samples analysed: 18

3 Data Processing

3.1 XCMS Peak Detection

Raw mzXML files were processed using XCMS (v4.10.1):

  • Peak detection: CentWave algorithm
  • Retention time correction: LOESS smoothing, span = 0.6, 2 rounds
  • Peak grouping: density method, bandwidth = 10
  • Missing peak filling: fillPeaks()
  • Adduct and isotope annotation: CAMERA (positive mode)
  • Total features detected: 11724

3.2 Preprocessing Pipeline

Table 2: Step-by-step preprocessing summary
Step Result
Raw XCMS features 11724 features detected
Zeros converted to NA 5991 zeros replaced with NA
50% per-group presence filter 11724 features retained — 0 removed
Half-minimum imputation All missing values replaced — 0 NAs remaining
TIC normalisation All 18 samples normalised to equal total ion count
Log2 transformation Log-normal distribution corrected
Autoscaling (z-score per feature) Mean = 0, SD = 1 per feature across samples

4 Quality Control

4.1 Lot Effect Assessment

Figure 1: PCA coloured by biological lot. Complete overlap between Lot1 and Lot2 confirms no significant technical batch effect between the two independent biological replicates.

Figure 1: PCA coloured by biological lot. Complete overlap between Lot1 and Lot2 confirms no significant technical batch effect between the two independent biological replicates.

Lot1 and Lot2 samples overlapped completely in PCA space confirming no significant technical effect between the two independent biological lots. Lot was therefore not included as a covariate in statistical models.

4.2 Biological Group Separation

Figure 2: PCA scores plot coloured by biological group. PC1 separates BNC5 from RSw-containing groups. PC2 separates BNC5RSw from RSw alone. All three groups show complete separation with no overlap between 95% confidence ellipses.

Figure 2: PCA scores plot coloured by biological group. PC1 separates BNC5 from RSw-containing groups. PC2 separates BNC5RSw from RSw alone. All three groups show complete separation with no overlap between 95% confidence ellipses.

PC1 (37.5% of variance) clearly separated BNC5 from the RSw-containing conditions. PC2 (17.5%) distinguished BNC5RSw from RSw alone. The positioning of BNC5RSw between BNC5 and RSw along PC1 reflects its composite metabolome containing both BNC5-derived and RSw-derived metabolites.

5 Statistical Analysis

5.1 ANOVA Results

7744 of 11724 features showed significant variation across the three groups after Benjamini-Hochberg correction (p < 0.05).

Table 3: Top 15 features by ANOVA significance (BH-corrected)
Feature Raw p-value BH-adjusted p Significant
M209.1854T2459.65 2.96e-24 3.47e-20 TRUE
M244.2218T2458.95 3.78e-23 2.21e-19 TRUE
M605.5372T2577.56 9.90e-23 3.16e-19 TRUE
M1486.0275T2464.02 1.08e-22 3.16e-19 TRUE
M736.5008T2412.35 5.95e-22 1.40e-18 TRUE
M688.4819T2453.61 2.07e-21 3.62e-18 TRUE
M1449.0594T2560.52 2.16e-21 3.62e-18 TRUE
M474.2553T1781.54 2.80e-21 4.10e-18 TRUE
M689.4844T2453.93 6.85e-21 8.92e-18 TRUE
M210.1887T2459.57 1.18e-20 1.38e-17 TRUE
M737.5029T2412.39 1.84e-20 1.96e-17 TRUE
M714.4957T2483.19 5.11e-20 5.00e-17 TRUE
M226.2119T2458.34 6.47e-20 5.84e-17 TRUE
M371.2206T2802.03 9.04e-20 7.57e-17 TRUE
M1145.9098T2565.67 9.99e-20 7.81e-17 TRUE

5.2 RSw Background Filter

To identify metabolites specifically produced by BNC5 in response to RSw, a three-step filter was applied:

Filter Step Count
Features elevated in BNC5RSw vs BNC5 (log2FC > 0.58, p_BH < 0.05) 2847
Removed — also elevated in RSw background 1314
True BNC5 response features 1533
Features reduced in BNC5RSw vs BNC5 1559

5.3 Volcano Plot

Figure 3: Volcano plot of BNC5RSw vs BNC5. Red = true BNC5 response features. Orange = RSw background features removed by filter. Blue = features reduced in BNC5RSw. Grey = not significant.

Figure 3: Volcano plot of BNC5RSw vs BNC5. Red = true BNC5 response features. Orange = RSw background features removed by filter. Blue = features reduced in BNC5RSw. Grey = not significant.

6 Multivariate Analysis — PLS-DA

Figure 4: PLS-DA scores plot using 1,533 true BNC5 response features. Complete separation of all three groups confirms biological specificity of the identified response metabolome.

Figure 4: PLS-DA scores plot using 1,533 true BNC5 response features. Complete separation of all three groups confirms biological specificity of the identified response metabolome.

7 Top BNC5 Response Features

Table 4: Top 20 true BNC5 response features ranked by VIP score
Feature m/z RT (min) Log2FC Fold Change p_BH VIP Score
M209.1854T2459.65 209.1854 40.99 4.173 18.04 1.91e-12 1.126
M210.1887T2459.57 210.1887 40.99 2.572 5.94 2.13e-11 1.126
M688.4819T2453.61 688.4819 40.89 5.806 55.96 9.48e-11 1.126
M689.4844T2453.93 689.4844 40.90 5.947 61.68 9.89e-10 1.126
M244.2218T2458.95 244.2218 40.98 4.669 25.44 1.39e-08 1.126
M605.5372T2577.56 605.5372 42.96 3.212 9.27 2.71e-08 1.126
M1449.0594T2560.52 1449.0594 42.68 12.727 6780.91 6.00e-09 1.125
M226.2119T2458.34 226.2119 40.97 2.322 5.00 7.56e-08 1.125
M578.4196T2801.03 578.4196 46.68 8.412 340.59 1.38e-07 1.125
M371.2206T2802.03 371.2206 46.70 9.122 557.13 1.40e-07 1.125
M714.4957T2483.19 714.4957 41.39 3.523 11.50 3.04e-07 1.125
M702.4961T2487.28 702.4961 41.45 4.061 16.69 3.32e-07 1.125
M537.3803T2801.03 537.3803 46.68 7.354 163.59 6.01e-07 1.125
M1486.0275T2464.02 1486.0275 41.07 6.038 65.71 2.13e-11 1.124
M1511.0407T2468.31 1511.0407 41.14 6.466 88.39 8.92e-10 1.124
M267.2841T1652.94 267.2841 27.55 10.591 1542.72 9.31e-10 1.124
M481.3070T1938.18 481.3070 32.30 4.704 26.07 1.70e-09 1.124
M736.5008T2412.35 736.5008 40.21 4.758 27.06 3.85e-09 1.124
M1145.9098T2565.67 1145.9098 42.76 9.411 680.64 1.94e-08 1.124
M759.5611T2610.96 759.5611 43.52 6.334 80.70 1.56e-07 1.124

8 Conclusions

Untargeted positive-mode LC-MS metabolomics of B. amyloliquefaciens BNC5 pellet extracts identified 1533 features specifically elevated in BNC5 upon exposure to R. solanacearum whole cells, after rigorous removal of RSw background metabolites (n = 1314).

A further 1559 features were significantly reduced in BNC5RSw compared to BNC5 alone, indicating active metabolic reprogramming in response to pathogen challenge — consistent with resource reallocation toward defensive compound biosynthesis.

The complete absence of lot effect confirms biological reproducibility across two independent experimental preparations. PCA and PLS-DA both confirmed complete metabolic separation between all three groups, validating the experimental design and computational workflow.

Recommended next steps:

  1. Putative annotation of the 1533 true BNC5 response features using accurate mass database searching (HMDB, METLIN, LipidMaps) and MS/MS fragmentation confirmation
  2. Supernatant fraction analysis using the identical pipeline to characterise secreted metabolites
  3. Negative ionisation mode analysis for complementary lipid and organic acid coverage

Analysis performed by Augustine Onuh, PhD
Department of Chemistry, Chulalongkorn University, Bangkok, Thailand
Generated: 2026-08-23 | R 4.6.0 | xcms v4.10.1