Image-seq: spatially resolved single-cell sequencing guided by in situ and in vivo imaging

Understanding tissue function requires knowledge of cellular organization. While single-cell RNA sequencing (scRNA-seq) provides insights into cellular heterogeneity, it lacks spatial context. Spatial transcriptomics aims to bridge this gap, but existing methods have limitations. Techniques like MERFISH, seqFISH, and Slide-seq link gene expression to spatial location but are often limited to in vitro studies or require tissue sectioning, hindering the analysis of complex 3D tissues. Furthermore, these techniques typically provide only static snapshots, making the study of dynamic cellular processes challenging. The current lack of integration with in vivo imaging further limits their application to dynamic processes. This paper introduces Image-seq, a novel platform that combines image-guided cell isolation with scRNA-seq, enabling spatially resolved transcriptomic analysis of cells within their 3D tissue context and over time. The bone marrow, a complex tissue crucial for hematopoiesis and a frequent site of leukemia development and metastasis, serves as a model system to demonstrate the capabilities of Image-seq. The high sensitivity and versatility of Image-seq allow the study of rare cell populations, such as leukemia cells in the early stages of development, providing crucial information that cannot be obtained through traditional methods. The ability to track and isolate individual cells in vivo and then analyze their gene expression profile promises to significantly advance our understanding of complex biological systems.

Existing spatial transcriptomics methods offer varying approaches to link gene expression to location. MERFISH, seqFISH, and Slide-seq achieve high-resolution spatial mapping but are often restricted to in vitro settings or require tissue sectioning, limiting their application to complex 3D structures like the bone marrow. Niche-seq represents a previous attempt to address spatial resolution in vivo but falls short in the depth of transcriptomic information. Many of these methods are limited by section thickness or reliance on indirect methods like pseudo-time analysis to infer temporal dynamics. The need for a technology enabling image-guided cell isolation, preserving cell viability, and providing high-sensitivity RNA sequencing readouts, particularly for rare cell populations within complex tissue microenvironments, is clearly evident in the literature. The lack of in vivo imaging integration in previous approaches underscores the unmet need addressed by Image-seq.

Image-seq integrates multiphoton microscopy with laser micromachining for image-guided cell isolation and scRNA-seq. A multiphoton microscope with two optical paths (one for imaging, one for ablation) is used. A femtosecond laser, operating at 5 MHz repetition frequency, is used for both imaging (1-2 nJ pulse energy) and plasma-mediated laser ablation (10-15 nJ pulse energy). A small channel (~50 x 100 µm) is created in the tissue to access the target cells. A micropipette is then inserted, and cells are aspirated under image guidance. The aspirated cells form a single-cell suspension, which is then processed using either high-throughput droplet-based sequencing (10x Genomics) or SMARTseq-v4 for individual cell profiling. The entire procedure, from imaging to cell aspiration, takes approximately 20 minutes per location. For the bone marrow studies, samples were isolated from β-actin-GFP mice, enabling direct visualization and targeting. Transcardial perfusion minimized red blood cell contamination, simplifying sample processing. Flow cytometry was used to validate cell population proportions in micropipette samples, and both 10x Genomics and SMARTseq-v4 were used to profile cells. AML progression was tracked using intravital microscopy in AML mouse models, where the spatial heterogeneity of leukemia cell expansion was observed. Cells were isolated from high and low-burden regions to determine spatial differences in gene expression. In the early leukemia progression study, intravital microscopy enabled classification of AML cells into proliferating (P), intermediate (IM), and non-proliferating (NP) populations based on cell division dynamics. Image-seq was used to isolate these cells, and SMARTseq-v4 was employed to minimize cell loss. The DPP4 protein was further analyzed using flow cytometry and in vivo imaging.

Image-seq successfully isolates viable cells from specific spatial locations within intact tissues, including bone marrow, both in situ and in vivo. The technique shows high sensitivity, achieving broad transcript coverage comparable to or exceeding conventional scRNA-seq protocols, even for rare cell populations (<0.01% leukemic burden). Analysis of bone marrow samples revealed no significant differences in cell population proportions between Image-seq and whole bone marrow preparations when aggregated, despite the spatial selectivity of Image-seq. In the AML model, Image-seq revealed spatial heterogeneity in early leukemia expansion. The study identified DPP4 as a key gene upregulated in proliferating AML cells within specific bone marrow microenvironments (M-type cavities). DPP4 expression was significantly higher in proliferating (P) AML cells compared to non-proliferating (NP) cells, both in vivo and in a secondary analysis of human AML datasets. Importantly, DPP4 expression was not observed in AML cells cultured in vitro, highlighting its dependence on the in vivo microenvironment. Further investigation revealed that co-culture of AML cells with specific bone marrow stromal and osteoblast precursor cells induced DPP4 expression, suggesting a role for niche-specific signals. Analysis of DPP4 correlated genes across both murine and human AML datasets shows a strong enrichment of cell cycle associated terms.

Image-seq provides a crucial advancement in spatial transcriptomics by overcoming limitations of existing methods. The ability to perform image-guided cell isolation in situ and in vivo, coupled with high-sensitivity scRNA-seq, provides unprecedented spatial and temporal resolution. The identification of DPP4 as a marker of proliferative AML cells within specific bone marrow microenvironments is a significant finding, suggesting potential therapeutic targets. The in vivo-specific expression of DPP4 further underscores the importance of studying cellular dynamics and gene expression within their native microenvironment. This work provides a powerful new tool to dissect the complex interplay between cells and their microenvironment in a wide range of biological contexts.

Image-seq successfully integrates in vivo/in situ imaging with scRNA-seq, offering superior spatial and temporal resolution for studying cellular processes. The identification of DPP4 as a marker of proliferative AML cells opens new avenues for therapeutic intervention. Future studies could focus on exploring the mechanisms by which bone marrow microenvironments regulate DPP4 expression and its effects on AML progression. Furthermore, the method can be applied to study a broader range of tissues and diseases, paving the way for deeper understanding of cellular organization and dynamics within their natural context.

The current Image-seq platform has a trade-off between spatial resolution and throughput. While high spatial resolution is achievable when isolating small numbers of cells, this reduces throughput. Future work should aim to improve throughput while maintaining high spatial resolution. The studies were largely limited to AML mouse models and further validation in human samples will be crucial to ensure clinical relevance. While the study suggests a clear link between DPP4 and AML proliferation, further investigation is needed to fully elucidate the underlying mechanisms and its clinical significance.