Virtually all cellular processes are influenced by environmental context, enacted in part through binding interactions between secreted ligands and their cognate receptors expressed on cell surfaces. Receptor-ligand (R-L) -mediated communication events are critical to establishment of cellular heterogeneity in healthy tissues, and may also present opportunities for therapeutic intervention when such interactions drive disease phenotypes. To maximize the potential impact of any investigation into these events, approaches should thus be applicable to suggest key communication mechanisms in normal tissue homeostasis, or lack thereof.
Indeed, a comprehensive data source profiling cell-cell communication across broad tissue states already exists: the recent surge in genomics technologies, particularly those capturing single cell measurements, has provided a massive amount of (yet un-tapped) information speaking to R-L usage across the phenotypic landscape. We propose that further development of computational approaches to dissect such events from single cell RNA sequencing (scRNA-seq) data would be invaluable in deriving mechanistic insights of cancer, embryonic development, and beyond. The Human Cell Atlas project (1) has already collected millions of cells from dozens of tissues, including the epithelial components of each organ along with interacting immune and stromal cells; it is my belief that few richer data sources exist today for comprehensive evaluation of cell-cell communication events in human biology.
Several approaches exist to infer cell-cell interactions via expression of cognate R-L pairs across cell types from scRNA-seq data, defining significant interactions with respect to a null model of random R-L expression across cells (2, 3). However, their accuracy is impacted by the limited capture rate for individual genes (particularly stable cell-surface proteins). Some methods address weak signal by incorporating gene expression downstream of receptor binding, but these approaches rely heavily on signaling pathways often defined in vitro and without regard to context-specific, cell-intrinsic signaling.
Here, we propose an approach (4) rooted in modules of inflammation-associated genes: each cell can receive signals based on its expressed receptors and send signals based on its expressed ligands. Specifically, “communication modules” are sets of receptor or ligand genes that tend to be mutually expressed in the same populations, and hence summarize the possible incoming and outgoing communication for a particular cell-state. To identify these patterns, our method—called Calligraphy—first builds a co-expression network of communication genes, from which robust inferences can be made across sub-populations representing coherent inflammatory programs. Prior knowledge of R-L relationships then inform the communication potential across cell-states based on their relative module usage.
To demonstrate the power of this approach, we have applied it to scRNA-seq data from pre-cancerous pancreatic cells harboring initiating mutations in the Kras proto-oncogene, actively undergoing transformation following an inflammatory event. Using Calligraphy, we first obtained 7 communication modules comprising genes that are co-expressed across the pancreatic epithelium (Figure 1A). Mapping average expression of these genes back onto the epithelium, we found that most cells express a single dominant module, making it possible to annotate cells by their corresponding module. Strikingly, states defined solely by communication gene expression coincide with those identified by clustering the entire transcriptome, suggesting a role for cell-cell communication in defining global cell-state heterogeneity (Figure 1B).
Applying Calligraphy to immune data collected in parallel identified consistently structured communication modules, many containing known regulators of pancreatic tumorigenesis. For example, myeloid module 20 is mainly expressed in macrophages and includes the type 2 IL-4 receptor and CSF2 receptor genes as well as Mif (5) and Cxcl1 (6), which have known roles in advanced cancer. Within the lymphoid compartment, module 8 is highly expressed in ILC2 and Treg cells; these cells express the receptor for IL-33, a ligand that accelerates the formation of mucinous pre-cancerous lesions (7).
With these modules in hand, Calligraphy nominates potential cell-cell interactions that drive the process of tumorigenesis, assuming that two modules potentially interact across cell subsets if they are significantly enriched in the number of shared cognate R-L pairs spanning them (Figure 2). We probed the Calligraphy network to systematically enumerate cycles involving any epithelial or immune subsets, and identified only one putative feedback loop in the system involving IL-33. Validation of this loop with in vivo IL-33 perturbation revealed far-reaching impacts of IL-33 signaling which are accurately predicted by Calligraphy.
The above results demonstrate the utility of applying a systems-level quantitative approach to predict critical cell-cell communication events. We expect the implications of this method can be far-reaching if applied to the numerous biological contexts now routinely profiled with scRNA-seq. Future work will involve application of Calligraphy to these new datasets, the output of which holds promise to nominate candidate cell-cell interactions spanning species, tissues, and cell-states.
1. A. Regev, S. A. Teichmann, E. S. Lander, I. Amit, C. Benoist, E. Birney, B. Bodenmiller, P. Campbell, P. Carninci, M. Clatworthy, H. Clevers, B. Deplancke, I. Dunham, J. Eberwine, R. Eils, W. Enard, A. Farmer, L. Fugger, B. Göttgens, N. Hacohen, M. Haniffa, M. Hemberg, S. Kim, P. Klenerman, A. Kriegstein, E. Lein, S. Linnarsson, E. Lundberg, J. Lundeberg, P. Majumder, J. C. Marioni, M. Merad, M. Mhlanga, M. Nawijn, M. Netea, G. Nolan, D. Pe’er, A. Phillipakis, C. P. Ponting, S. Quake, W. Reik, O. Rozenblatt-Rosen, J. Sanes, R. Satija, T. N. Schumacher, A. Shalek, E. Shapiro, P. Sharma, J. W. Shin, O. Stegle, M. Stratton, M. J. T. Stubbington, F. J. Theis, M. Uhlen, A. van Oudenaarden, A. Wagner, F. Watt, J. Weissman, B. Wold, R. Xavier, N. Yosef, Human Cell Atlas Meeting Participants, The Human Cell Atlas. Elife. 6 (2017), doi:10.7554/eLife.27041.
2. M. Efremova, M. Vento-Tormo, S. A. Teichmann, R. Vento-Tormo, CellPhoneDB: inferring cell-cell communication from combined expression of multi-subunit ligand-receptor complexes. Nat. Protoc. 15, 1484–1506 (2020).
3. R. Browaeys, W. Saelens, Y. Saeys, NicheNet: modeling intercellular communication by linking ligands to target genes. Nat. Methods. 17, 159–162 (2020).
4. C. Burdziak, D. Alonso-Curbelo, T. Walle, F. M. Barriga, J. Reyes, Y. Xie, Z. Zhao, C. J. Zhao, H.-A. Chen, O. Chaudhary, I. Masilionis, Z.-N. Choo, V. Gao, W. Luan, A. Wuest, Y.-J. Ho, Y. Wei, D. Quail, R. Koche, L. Mazutis, T. Nawy, R. Chaligné, S. W. Lowe, D. Pe’er, Epigenetic plasticity cooperates with emergent cell-cell interactions to drive neoplastic tissue remodeling in the pancreas. bioRxiv (2022), , doi:10.1101/2022.07.26.501417.
5. N. Funamizu, C. Hu, C. Lacy, A. Schetter, G. Zhang, P. He, J. Gaedcke, M. B. Ghadimi, T. Ried, H. G. Yfantis, D. H. Lee, J. Subleski, T. Chan, J. M. Weiss, T. C. Back, K. Yanaga, N. Hanna, H. R. Alexander, A. Maitra, S. P. Hussain, Macrophage migration inhibitory factor induces epithelial to mesenchymal transition, enhances tumor aggressiveness and predicts clinical outcome in resected pancreatic ductal adenocarcinoma. Int. J. Cancer. 132, 785–794 (2013).
6. L. Seifert, G. Werba, S. Tiwari, N. N. Giao Ly, S. Alothman, D. Alqunaibit, A. Avanzi, R. Barilla, D. Daley, S. H. Greco, A. Torres-Hernandez, M. Pergamo, A. Ochi, C. P. Zambirinis, M. Pansari, M. Rendon, D. Tippens, M. Hundeyin, V. R. Mani, C. Hajdu, D. Engle, G. Miller, The necrosome promotes pancreatic oncogenesis via CXCL1 and Mincle-induced immune suppression. Nature. 532, 245–249 (2016).
7. D. Alonso-Curbelo, Y.-J. Ho, C. Burdziak, J. L. V. Maag, J. P. Morris 4th, R. Chandwani, H.-A. Chen, K. M. Tsanov, F. M. Barriga, W. Luan, N. Tasdemir, G. Livshits, E. Azizi, J. Chun, J. E. Wilkinson, L. Mazutis, S. D. Leach, R. Koche, D. Pe’er, S. W. Lowe, A gene-environment-induced epigenetic program initiates tumorigenesis. Nature. 590, 642–648 (2021).