TL;DR
This paper introduces "Spatial Meta-transcriptomics" (SMT), a novel computational pipeline that allows researchers to simultaneously map the spatial distribution of microbial communities and host gene expression within tissue samples. This breakthrough enables a deeper understanding of host-microbe interactions at an unprecedented spatial resolution, opening new avenues for studying diseases and biological processes.
What Problem Does This Paper Address?
Traditional microbiome studies, while powerful, typically lack spatial resolution, treating microbial communities as a bulk entity. This makes it difficult to understand how specific microbes interact with specific host cells or tissues in their native environment. Spatial transcriptomics, on the other hand, excels at mapping host gene expression in situ but has historically been unable to capture microbial signals. This paper addresses the critical gap of integrating these two powerful approaches, aiming to provide a spatially resolved view of both host and microbial landscapes within the same tissue. This is particularly relevant for organs like the intestines, which harbor a dense and functionally important microbial community.
Methods
The core innovation is an analysis pipeline designed to extract microbial sequences from existing spatial transcriptomic data. This pipeline works by:
- Extracting Unmapped Reads: After standard spatial transcriptomic analysis, reads that don't map to the host genome are isolated.
- Taxonomic Assignment: These unmapped reads are then aligned against a microbial reference database (NCBI NT) to assign taxonomic labels (e.g., species, genus, family).
- UMI-based Abundance Matrix: Reads originating from the same Unique Molecular Identifier (UMI) are grouped, and a consensus taxonomic assignment is made for that UMI. This allows for the generation of a spatial microbial abundance matrix, similar to the host gene expression matrix produced by standard spatial transcriptomics.
The authors also employed Fluorescence In Situ Hybridization (FISH) for validation and bulk RNA sequencing to compare their SMT results with a more traditional microbial profiling method.
Datasets
The study utilized both human and murine intestinal samples.
- Human Samples: Colon samples from two colorectal cancer patients, including both tumor and adjacent normal tissue.
- Murine Samples: Small and large intestine cross-sections from three mice.
These datasets were generated using the 10x Genomics Visium spatial transcriptomics platform. The authors also used publicly available single-cell sequencing data for comparison.
Key Insights and Ideas
- Microbial Presence in Unexpected Places: The SMT approach revealed microbial signals not only in the intestinal lumen but also within tumor tissues, suggesting deeper penetration and potential interactions with host cells in compromised barrier environments.
- Validation of SMT: The spatial patterns of microbial abundance observed with SMT were consistent with expectations (e.g., higher microbial load in the lumen) and were further validated by FISH, lending strong credibility to the method.
- Host-Microbe Correlation: SMT enabled the identification of spatial correlations between specific microbial taxa and host gene expression. For instance, they found correlations between Porphyromonas and immune defense genes like Saa1 in the small intestine, and Helicobacter with Dmnt1 in the large intestine.
- Detection of Viral Infections: The pipeline's ability to capture viral sequences alongside bacterial and fungal ones was demonstrated by the identification of Cytomegalovirus (CMV) infection within tumor spots in a human CRC dataset, allowing for detailed spatial mapping of the infection.
- Bias Assessment: The authors carefully assessed potential biases introduced by the spatial transcriptomics protocol (e.g., poly-A selection) and found that while some microbial sequences might be depleted, the method still captures significant and meaningful microbial signals, with ribosomal RNAs being a major contributor.
Key Results
- The SMT pipeline successfully generated spatial microbial abundance matrices alongside host gene expression matrices for both human and murine intestinal samples.
- The percentage of microbial sequences varied significantly, from 7.32 x 10^-7 in human samples to 2.26 x 10^-3 in murine samples.
- SMT results showed strong correlations with bulk RNA sequencing data for both host genes (R=0.91) and microbial abundances at the family level (R=0.83), validating the approach.
- In one human CRC dataset, viral sequences (Cytomegalovirus) accounted for up to 3.4% of reads in tumor spots.
- The study identified specific spatial colocalizations of host genes and microbial genera, such as Porphyromonas with Saa1 and Helicobacter with Dmnt1.
Limitations and Future Work
- Capture Efficiency: The authors acknowledge that the permeabilization step in spatial transcriptomics might not be ideal for capturing all microbial sequences, potentially leading to some biases. However, they argue that the captured signals are still meaningful and validated.
- Sensitivity: While the FISH results suggest the current SMT method is sensitive, future work aims to further increase sensitivity and taxonomic resolution.
- Future Directions: The authors envision that by capturing more sequences or using specifically designed capture oligos (e.g., for 16S rRNA), the SMT approach could be further enhanced. This technology has the potential to revolutionize the study of host-microbe interactions in various tissues, including sterile environments and the investigation of microbial translocation to remote sites.
Summary updated