Tractable mouse TNBC models capture the heterogeneous tumor immune microenvironment and adaptation to PD-L1 blockade

Triple-negative breast cancer (TNBC) patients exhibit variable responses to programmed death (PD)-ligand (L)1 blockade, largely determined by the 'hot' versus 'cold' state of the tumor immune microenvironment (TIME). We here characterized 9 mouse TNBC models, relying on intraductal mammary gland inoculation of established mouse TNBC cell lines, with a heterogeneous TIME to study anti-PD-L1 resistance mechanisms. Complementary in vitro and in vivo screening classified the 4T1-hot-based model, a highly inflamed control through its immunogenic luciferase tag expression compared to the untagged 4T1-cold-based model, as displaying the 'hottest' TIME. However, both 4T1-based counterparts did not respond to anti-PD-L1, which was attributed to their immunosuppressive myeloid cell content as well as upregulation of cancer-associated fibroblasts in the 4T1-hot and high PD-L1-expressing CXCL10+ tumor-associated macrophages in 4T1-cold primary tumors. These anti-PD-L1 adaptation mechanisms across TIME states as captured by mouse TNBC models highlight specific cellular targets for future studies. Overall design: RNA was isolated from primary tumors of all TNBC mouse models (D2A1, EMT6, 4T1-cold, AT3, 4T1-hot, 66CL4, PY230, MET1 and MVT1) at 2 w p.i and healthy mammary glands from Matrigel-inoculated BALB/C, C57BL/6 and FVB/N mice using Rneasy Mini Kit (QiAgen, Valencia, CA, USA). All analyses were done on 3 samples for all primary tumors and healthy mammary glands, except n=4 for PY230 and AT3 primary tumors. Bulk RNA sequencing was done in collaboration with Azenta Life Sciences and performed on Illumina Novaseq system (paired end, length of 150 bp), aiming for 10 million reads per sample. Differential expression of 2 w primary tumors was obtained for each model comparing to their respective healthy control mammary gland (BALB/c vs 4T1-hot, 4T1-cold, EMT6, D2A1 and 66CL4 tumors, C57BL/6 vs Py230 and AT3 tumors, FVB/N vs MVT1 and MET1 tumors). Identifying hallmarks of cancer for each model and comparing these models to eachother based on log2-fold changes in expressed genes using selection of different gene sets.