Automatic metabolome profiling of bacterial colony heterogeneity by multimodal imaging with mass spectrometry and microscopy

Introduction
Mass Spectrometry Imaging (MSI) is a method of growing interest for the in-situ study of metabolites produced during bacterial colony interactions on hard surfaces. However, this type of analysis is often limited to one interaction per MS image or very few interactions. With this practice, the heterogeneity of different bacteria, i.e. the diversity in metabolites expression within the same bacterial strain, is not taken into account. Here, we propose a new informatic method that allows to study the molecular expression of multiple bacterial colonies from a single MS image by combining light microscopy and MALDI mass spectrometry imaging. The method automatizes the detection of different colonies and attributes signal from every detected metabolite to each colony.
Methods
Bacteria are grown on a thin layer of agar directly on an ITO (indium tin oxide) plate. A bright field microscopic image of the plate is taken before the MALDI preparation of the plate (pre-MALDI image). After MALDI MS acquisition of the plate a new bright field image of the plate is taken (post-MALDI image). A cross modality image registration is performed according to the MS, pre-MALDI and post-MALDI image. An image segmentation pipeline allows to determine the position of every detectable object on the plate. Then, the MS signal corresponding to each of those objects is estimated. This method generates a matrix of MS signal expression where rows correspond to the objects and columns to their metabolites.
Preliminary Data
We have applied our method on an agar coculture of Bacillus velezensis GA1 and Pseudomonas sp. CMR12a. Both strains are biocontrol agents that produce lipopeptides. The bacteria are inoculated on the opposite side of an ITO plate in such a way that the middle of the plate corresponds to the interaction between the 2 different strains. The bacteria are grown during 24 hours forming multiple micro-colonies of around 300 µm of diameter. Control plates are generated by inoculating each strain alone. The MSI of the plate is acquired on a MALDI FT-ICR-MS (SolariX XR 9.4T, Bruker) identifying different families of lipids such as phosphatidylethanolamine (PE), phosphatidylglycerol (PG) and lipopeptides such as surfactins, orfamides and sessilins. The MS signal of each micro-colony is estimated according to our method and the corresponding image of each colony is extracted from the pre-MALDI image. The method detects multiple objects on the pre-MALDI image which is then filtered to isolate CMR12a colonies. The removed objects correspond to GA1 colonies and artefacts with not enough signal to be analyzed. It is then possible to investigate the heterogeneity of the colonies by applying unsupervised clustering algorithms (hierarchical clustering). Statistical analysis is used to detect specific signals of those clusters. In this experiment, our method highlights a sub-cluster of CMR12a which is characterized by an over expression of particular PG lipids and an under expression of sessilins and orfamides. The colonies corresponding to this sub-cluster seems to be located closer to the interaction region with GA1 compared to the other CMR12a colonies. Moreover, the method highlights multiple CMR12a mutants previously identified as mutants that have lost a genomic island which over expressed PE lipids and do not express sessilins. Further work will focus on in vivo analysis of bacteria colonizing plant’s roots.
Novel Aspect
Informatic method allowing the automatic detection and the study of multiple bacterial micro-colonies by their MS and microscope image.