Engineering a spatiotemporal macrophage circuit via STING phase separation to override immune suppression in pancreatic cancer

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies, largely due to its highly immunosuppressive tumor microenvironment (TME), which fuels metastasis and resistance to immunotherapy. Through comprehensive analysis of single-cell RNA sequencing (scRNA-seq) datasets, we identified multiple heterogeneous tumor-associated macrophage (TAMs) subpopulations as key regulators of PDAC progression, which co-express MRC1 and exert their effects by actively suppressing anti-tumor immune responses. To overcome this barrier, we developed a spatiotemporal macrophage reprogramming platform that leverages STING phase separation to reprogram TAM plasticity and reshape the immune landscape. This system, PMMB, integrates a CSF-1R inhibitor and a STING agonist within a macrophage-mimetic nanostructure, enabling sequential, controlled reprogramming of TAMs. By leveraging STING phase separation, PMMB stabilizes TAMs in an anti-tumor CD80⁺ phenotype while preventing excessive inflammation, achieving durable immune activation. In preclinical models, PMMB not only suppresses both primary and metastatic PDAC but also enhances CD8⁺ T cells infiltration, reinvigorates anti-PD-1 therapy responses, and mitigates immune exhaustion. These findings establish spatiotemporal macrophage circuit engineering via STING phase separation as a novel cross-scale strategy to override PDAC’s immune barriers and drive next-generation macrophage-targeted immunotherapy. This study paves the way for rationally designed, precision macrophage modulation strategies in solid tumors.

Methods
Single-cell RNA sequencing of PDAC tumors in mice

To create a subcutaneous xenograft PDAC model, 5×106 KPC cells were administered via injection into the right flank of male C57BL/6 mice. PMMB was injected into the tail vein of mice on days 0, 3, 6, 9, and 12. On Day 14, the tumors from the mice were collected for scRNA-seq.

Tumors were enzymatically dissociated into single-cell suspensions. After removal of dead cells, viable cells (>80%) were resuspended in PBS with 0.5% BSA and processed for scRNA-seq. Single-cell libraries were generated using a droplet-based platform according to the manufacturer’s protocol, including reverse transcription, cDNA amplification, and library indexing.

Libraries were sequenced on an Illumina NovaSeq platform, yielding ~200,000 reads per cell with a median detection of ~2,000–2,500 genes. Raw reads were filtered with fastp (4), aligned to the mouse mm10 genome, and processed to generate gene-barcode matrices for downstream analysis in Seurat.

Bulk RNA sequencing

Total RNA was isolated from PMMB- and PBS-treated MRC1⁺ macrophages (n = 3 per group) and PDAC mouse tumor tissues using TRIzol reagent (Invitrogen) following the manufacturer’s instructions. RNA quality was confirmed using NanoDrop and Agilent Bioanalyzer (RIN > 7.0, OD260/280 > 1.8, total RNA > 1 μg). Libraries were constructed with the Illumina Stranded mRNA Prep Kit, including polyA selection, RNA fragmentation, cDNA synthesis, adapter ligation, and PCR amplification. Sequencing was performed on an Illumina NovaSeq 6000 platform (paired-end, 150 bp).

Raw reads were aligned to the mouse genome (mm10) or human genome (GRCh38) using STAR, and gene counts were generated with featureCounts. Differential expression analysis was performed with DESeq2 (|log₂ fold change| > 1, adjusted p < 0.05). DEGs were further subjected to GO, KEGG, and GSEA analyses to evaluate immune-related pathways, TAM-associated functions, and STING signaling.