TL;DR
This paper introduces Spatial Meta-Transcriptome (SMT), a novel computational pipeline that allows for the simultaneous analysis of host gene expression and microbial abundance within tissue samples at a spatial resolution. By adapting existing spatial transcriptomics (ST) data, SMT provides a powerful new tool to investigate host-microbe interactions in situ, opening doors to new biological discoveries.
What Problem Does This Paper Address?
Understanding the complex interplay between host organisms and their resident microbes is crucial for many areas of biology and medicine. Traditional methods for studying the microbiome (like 16S rRNA sequencing) typically provide information about the overall microbial composition but lack spatial context. Conversely, spatial transcriptomics (ST) excels at mapping gene expression within tissues but has historically struggled to capture microbial signals. This leaves a significant gap: how do we simultaneously study host gene expression and microbial distribution within the same tissue sample, at a fine spatial scale? This paper tackles this challenge by developing a method to extract and analyze microbial sequences directly from ST data, bridging the divide between host and microbial investigations.
Methods
The core of this work is the development of the Spatial Meta-Transcriptome (SMT) analysis pipeline. The process begins with standard ST data, which is typically generated by capturing mRNA from tissue sections. The SMT pipeline takes the raw sequencing reads and first filters out those that map to the host genome. The remaining "unmapped" reads are then aligned against a microbial reference database (NCBI NT) using BLAST. To handle conflicting taxonomic assignments for reads originating from the same Unique Molecular Identifier (UMI), SMT employs a strategy of majority voting or identifying the most recent common ancestor. This process generates a spatial microbial abundance matrix, which can then be integrated with the existing host gene expression matrix within a Seurat object.
To address potential biases and improve microbial capture, the authors also explored experimental modifications. They found that adding a 15-minute lysozyme treatment after tissue permeabilization significantly increased the capture of microbial transcripts, including those from Gram-positive and Gram-negative bacteria, as well as fungi, without negatively impacting host gene expression quality.
Datasets
The study utilized both human and murine intestinal tissue samples.
- Human samples: Colon tissue from two colorectal cancer patients, including both tumor and adjacent normal sites.
- Murine samples: Small and large intestine sections from three mice.
These samples were processed using the Visium Spatial Gene Expression platform. The authors also used positive control experiments with cultured Pseudomonas aeruginosa and Staphylococcus aureus to quantitatively assess SMT's capture efficiency and recall.
Key Insights and Ideas
- Bridging the Host-Microbe Divide: The most significant contribution is the ability to simultaneously profile host gene expression and microbial abundance in a spatially resolved manner. This allows for direct investigation of how host cellular states correlate with the presence and location of specific microbes.
- Leveraging Existing ST Data: SMT demonstrates that valuable microbial information can be extracted from existing ST datasets, making this powerful approach accessible without requiring entirely new experimental protocols for every study.
- Experimental Optimization is Key: The addition of a lysozyme treatment significantly enhances microbial capture, suggesting that optimizing sample preparation for microbial RNA is crucial for maximizing SMT's potential.
- Microbial Signals are Real: The observed spatial patterns of microbial distribution in the intestines (e.g., higher abundance in the lumen) and the correlation with known biological phenomena (like increased microbial penetration in tumors) strongly suggest that the detected microbial signals are not merely artifacts or contaminants.
Key Results
- SMT successfully generated a spatial microbial abundance matrix alongside the host expression matrix.
- The pipeline was applied to both human and murine intestinal sections, revealing spatial patterns of microbial distribution.
- Comparisons with alternative assays like FISH validated the spatial microbial abundance information obtained by SMT.
- The study identified host-microbe interactions at various spatial scales, including the localization of Cytomegalovirus within fibroblast-rich areas in a colorectal cancer patient and the potential migration of Clostridium across the intestinal wall.
- Positive control experiments demonstrated that SMT can achieve a recall of over 90% for specific bacteria, with an approximate capture rate of one bacteria sequence per two to four bacterial cells.
- The lysozyme treatment increased microbial detection by approximately 4-fold on average, with no significant negative impact on host gene expression quality.
Limitations and Future Work
- Sensitivity and Resolution: The authors acknowledge that in its current form, SMT may require a large number of samples for significant findings in humans, partly due to technical limitations like noise and resolution. Future improvements in ST technology, such as higher resolution and more sensitive capture methods, are expected to enhance SMT's capabilities.
- Microbial RNA Extraction: The current SMT framework primarily generates an abundance table. Future work aims to optimize microbial RNA extraction or develop specialized capture oligos to enable profiling of microbial transcriptomes.
- Taxonomic Resolution: While SMT provides taxonomic information, further research is ongoing to achieve higher resolution for bacteria identification.
- Preliminary Findings: Some findings, such as the migratory capabilities of Enterococcus and Staphylococcus, are noted as preliminary and require further validation with more SMT samples and orthogonal assays.
- Contaminant Filtering: While SMT includes methods for removing contaminants, the authors suggest that a known list of contaminants could be optionally used to further filter abundance matrices.
Summary updated