From transcriptional states to spatially validated tumor architecture: a multi-omics view of lung adenocarcinoma

Spatial transcriptomics has expanded our ability to identify cell types within tissues. However, understanding tumor biology requires more than composition alone. This article builds on findings from Takano et al. (Nature Communications, 2024) to examine how spatial organization, including gradients and compartmentalization, shapes interpretation of tumor systems.

Figure 1. Multi-omics overview of tumor analysis

RNA defines regions, but not certainty

Spatial transcriptomics enables the identification of distinct tumor regions based on gene expression patterns. In lung adenocarcinoma, transcriptional markers such as NKX2-1 and HNF4A can distinguish well-differentiated tumor areas from more malignant or mucinous regions.

Figure 2.Transcriptomic clustering separates tumor regions based on gene expression, providing an initial map of tumor heterogeneity

These classifications provide a useful map of tumor heterogeneity. However, they are fundamentally based on transcriptional inference. Transcriptomic signals captured by sequencing-based platforms such as Visium provide robust measurements of gene expression across tissue regions. However, these signals represent aggregated expression within each spatial unit and do not directly resolve the precise cellular location or functional activity of those transcripts.

As a result, tumor regions defined by RNA alone may represent potential biological states rather than fully validated ones.

Tumor heterogeneity is spatially structured

Tumor microenvironments are organized into spatially distinct regions, each associated with different biological behaviors. In this dataset, well-differentiated regions characterized by NKX2-1 expression are spatially separated from mucinous or more malignant regions marked by HNF4A.

Figure 3.Canonical marker genes distinguish well-differentiated and malignant tumor regions at the transcriptional level

Within these malignant regions, additional features emerge, including inflammatory signaling and the presence of cancer-associated fibroblasts. These fibroblast populations form structured compartments between tumor regions, contributing to the physical and functional organization of the tissue.

This spatial structure suggests that tumor progression is not uniform, but occurs across distinct yet connected regions within the same tissue.

Interpreting inferred biology

While transcriptomic data can identify these regions, interpreting their biological significance remains challenging. For example, immune-related genes such as IDO1 may be highly expressed in certain regions, suggesting immunosuppressive activity.

Figure 4. Spatial alignment of RNA and protein signals confirms that immunosuppressive pathways are localized within specific tumor regions

However, without validation at the protein level, it remains unclear whether these signals correspond to active biological processes or transcriptional noise.

In addition, spatial transcriptomics at lower resolution may not fully resolve the boundaries between tumor, stromal, and immune compartments, making it difficult to interpret how these populations interact.

Leveraging imaging-based transcriptomics, sequencing-based transcriptomics, and proteomics spatial data

A multi-omics spatial approach provides a way to address these limitations.

Sequencing-based transcriptomic data (Visium) can be used to identify major tumor regions and define transcriptional states. Proteomic data (PhenoCycler) can then be aligned to the same tissue to validate whether key signals, such as IDO1 expression, are present at the protein level in the same spatial locations.

Rather than relying solely on transcriptional signals, key biological activity can be evaluated by examining whether these signals are consistently observed at both the RNA and protein levels within the same spatial regions.

When these layers are aligned, key signals can be evaluated with greater confidence. In this dataset, IDO1 expression is consistently observed at both the RNA and protein levels within the same regions, supporting its role in localized immune suppression.

Figure 5. Spatial concordance between IDO1+ transcriptomic regions and protein distribution highlights localized immunosuppressive activity within the tissue

From inferred regions to validated tumor structure

By leveraging these layers, tumor regions can be interpreted with greater confidence.

Higher-resolution imaging-based spatial data (Xenium) further refines these findings by resolving individual cells. This allows clear separation of tumor, stromal, and immune populations, and reveals structured features such as cancer-associated fibroblasts forming boundaries between tumor regions.

Figure 6. High-resolution spatial data reveals cellular organization and structural boundaries between tumor compartments

At the same time, higher-resolution data reveals how these regions are organized at the cellular level, providing insight into how tumor progression, immune activity, and stromal structure interact within the tissue.

Toward spatially validated tumor biology

This approach shifts spatial analysis from identifying patterns to validating mechanisms.

Tumor heterogeneity is no longer defined only by gene expression, but by the alignment of molecular signals across RNA and protein, and by their spatial organization within the tissue.

As spatial datasets continue to expand across modalities, combining these layers will become essential for interpreting tumor microenvironments with both resolution and confidence.

References

Takano, Y., Suzuki, J., Nomura, K., Fujii, G., Zenkoh, J., Kawai, H., Yuta Kuze, Kashima, Y., Nagasawa, S., Nakamura, Y., Kojima, M., Katsuya Tsuchihara, Seki, M., Kanai, A., Matsubara, D., Takashi Kohno, Noguchi, M., Nakaya, A., Tsuboi, M., & Ishii, G. (2024). Spatially resolved gene

‌