ORCID

https://orcid.org/0000-0002-9886-4039

Date of Award

Summer 2025

Language

English

Embargo Period

7-23-2027

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Biological Sciences

Program

Biology

First Advisor

Alex Valm

Committee Members

Alex Valm, Gabriele Fuchs, Paolo Forni, Jean-Marc Ghigo

Keywords

microbiome, machine learning, multiplex spectral imaging, spatial structure and function, phylogenic resolution, gnotobiotic model

Subject Categories

Bioinformatics | Biology | Environmental Microbiology and Microbial Ecology | Laboratory and Basic Science Research | Life Sciences | Microbiology

Abstract

Advances in imaging technologies and computational tools are transforming our understanding of the spatial organization of microbial communities within the host. Here, we investigated the spatial structure of the gut microbiome and its functional relationship with host mucus phenotypes by integrating high-resolution imaging, machine learning-based spectral unmixing, and gnotobiotic zebrafish models to uncover the mechanistic principles underlying host–microbe interactions. We first reviewed recent developments in microbiome spatial ecology, highlighting two distinct yet complementary dimensions of spatial structure: biogeography, the distribution of microbes across anatomical regions, and architecture, their fine-scale organization within a niche. This framework underscores how spatial context governs microbial community function and host interaction.

Using axenic, conventional, and gnotobiotic zebrafish models, we found that specific microbial taxa, rather than total bacterial biomass, drive mucus abundance and spatial patterning in a tissue- and context-dependent manner. For example, colonization with a defined bacterial consortium was sufficient to restore mucus production and gut architecture in axenic fish, recapitulating conventional phenotypes despite variation in overall microbial load—supporting the concept of microbial functional sufficiency. However, we also observed inherent variability in colonization outcomes and host responses across gnotobiotic individuals, highlighting a limitation of current re-conventionalization approaches. These findings reinforce the need to validate microbial colonization in gnotobiotic experiments to improve reproducibility and biological interpretation in host–microbiota studies.

Building on this framework, we addressed a major technical challenge in microbial imaging: resolving species-level identity in densely labeled, multiplexed communities. To overcome the resolution limitations of conventional fluorescence imaging and 16S rRNA-based methods, we developed, Cross-hybridization Inference and Phylogenetic Resolution fluorescence in-situ hybridization (CIPHR-FISH), a novel machine learning-based spectral classification framework for species-level identification in multiplex-labeled microbial communities. This approach was developed to test our functional sufficiency hypothesis more directly by enabling species-level resolution of microbial spatial patterns and identifying core microbiota configurations associated with differential host outcomes. This method generates and captures ‘cross hybridization inference’ that capture excitation/emission properties, spectral cross talk and probe cross-reactivity, enabling pixel-level classification of microbial species. Applied to a synthetic zebrafish gut consortium, the framework achieved 100% specificity and high sensitivity, surpassing traditional linear unmixing techniques and setting a new benchmark for microbial imaging. We then applied this high-resolution image analysis pipeline to examine microbial identity and spatial organization in vivo in the zebrafish gut and identified Aeromonas sp. as the dominant taxon in the presence of 5 other co-colonizing taxa—consistent with 16S rRNA sequencing data from these communities.

Altogether, this work integrates microbiome imaging, machine learning, and host physiology and development to reveal how microbial spatial structure shapes host mucosal environments, advancing tools our ability to map, resolve, and interpret microbiota–host relationships with unprecedented resolution.

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Available for download on Friday, July 23, 2027

Share

COinS