Registration for #SCG2025 in the beautiful Stockholm Archipelago will open on Monday, April 28 at 15.00 CEST. Space is limited so don’t miss your registration!: registration.invajo.com/82ec3735-7cb...
Conference web page: conferences.weizmann.ac.il/SCG2025/

In summary, our research established the blueprint of tumor macrophages and defined Zeb2 as a master switch controlling the TAM program, with implications for clinical translation and TAM-targeted immunotherapy. (19/20)

Among the tumor models we tested, we used an orthotopic bladder cancer model and found that intravesical treatment with CpG-conjugated Zeb2 siRNA showed robust tumor control, with the potential to replace conventional BCG therapy. (18/20)

By design, CpG alone activates inflammatory pathways in TAMs - but adding Zeb2 silencing improved significantly TAM reprogramming and tumor control across different tumor models. (17/20)

Next, since Zeb2 is a transcription factor, we asked if it also controls these programs at the chromatin level. ATAC-seq revealed that Zeb2 KO causes massive enhancer remodeling—opening sites near anti-tumor genes while closing pro-tumor ones. (15/20)

To validate our proposed anti-tumor score in a human context, we compared it to the CXCL9:SPP1 TAM polarity score linked to patient survival (beautiful work by Bill et al.). The effect of each regulator on the activity scores correlated well, with Zeb2 topping the list. (13/20)

With this, we established a TAM functional network and proposed a quantitative reprogramming metric—an 'anti-tumor score': how much a perturbation upregulates anti-tumor and downregulates pro-tumor functions. Among known hits like Trem2, Zeb2 emerged as a key switch. (12/20)

Next, we went back to our network and asked how the different perturbations affect these TAM programs. This is useful. Why? Because a major reprogramming target would upregulate anti-tumor (e.g. IFN-response) and downregulate pro-tumor (e.g. immunosuppression) programs. (11/20)

To understand how the different perturbations reprogrammed TAMs, we grouped autocorrelated genes using MrVI’s perturbation-aware latent space. This yielded distinct gene modules with known roles in TAM biology, linked to either positive or negative tumor immunity. (10/20)

This AI-driven analysis enabled us to construct a perturbation network, where perturbations that evoke similar effects on the single-cell states are located in close proximity—and vice versa. (9/20)

To systematically resolve how candidate regulators impact macrophage functions we collaborated with the Yosef lab who recently developed MrVI, a deep generative model designed to dissect how sample-level covariates like gene perturbation affect gene expression in single cells. (8/20)

The first look at the data confirmed that the putative regulators were effectively targeted—the blue diagonal indicates reduced gene expression of the perturbed genes. But the data, as you can see, is complex, high-dimensional, and not easily interpretable. (7/20).

The screen yielded ~85,000 single-cell profiles across 120 perturbations in one very, very long experiment! (6/20)

To model TAMs in vitro, we tested different cytokines and found that IL-4–treated bone marrow-derived macrophages indeed best mimic pro-tumor TAMs. This was the basis for our arrayed CRISPR screen with scRNA-seq as the readout. (5/20)

To begin our journey to understand TAM regulation, we integrated single-cell RNA-seq datasets from human tumors and healthy tissues to define core TAM gene programs. From this, we nominated ~120 conserved candidate regulators for CRISPR perturbation. (4/20)

Macrophages are the immune system's Swiss Army knives. Tumors exploit their multifunctionality, polarizing TAMs to activate tumor-supporting functions.
Our goal was to find targets to reprogram TAMs into tumor-fighting immune cells. (2/20)