Monitoring PD-L1 in tumor macrophage fusion cells in blood correlates to PD-L1 checkpoint inhibitor responses in metastatic breast cancer

In metastatic breast cancer (mBC), anti-PD-L1/PD-1 immune checkpoint inhibitors (ICIs), e.g. pembrolizumab, are approved in a subpopulation of mBC patients (pts) with a PD-L1 combined positive score (CPS) ≥10, median progression-free survival (mPFS) of 9.7 months and median overall survival (mOS) of >24 months. However, 62% of pts have CPS <10, and may also benefit from ICIs (i.e. ≥1 CPS have mPFS=7.6 months vs 5.6 for chemotherapy). One hypothesis to why low PD-L1 pts respond to ICIs is dynamic PD-L1 upregulation after starting a new therapy, requiring a biomarker to monitor PD-L1 and subsequent ICI benefit. Recent studies have identified PD-L1 expressing myeloid cells that disseminate into the blood from primary tumors, tumor-macrophage fusion cells (TMFCs), which may predict ICI response. However, dynamic changes in TMFC PD-L1 during ICI and their relationship to patient response is unknown. In this study, we monitored PD-L1 expression in TMFCs during ICI treatment, compared to tumor CPS, in mBC pts to evaluate PFS & OS at 24 months. We conducted a prospective pilot study of n=43 pts with pathologically confirmed mBC prior to receiving anti-PD-L1 ICIs. CellSieve microfilters isolated TMFCs from 7.5ml peripheral blood samples at 4 time points, prior to start of PD-L1 ICI (T0) and at monthly timepoints (T1-T3) for 4 months after ICI induction. TMFCs were identified by enlarged cell size (>30 µm) and polyploid nucleus. Average PD-L1 expressions in TMFCs were categorized as negative/low or high. Pearson's correlation compared average TMFC PD-L1 to CPS PD-L1 from tissue. TMFC PD-L1 expression and CPS were compared to pts' PFS and OS by Cox proportional univariate/multivariate analysis at 24 months. 95.3% (n=41/43) of pts provided a T0 sample. 90.7% (n=39/43) of pts provided a T1 sample (~28 days after ICI). 67.4% (n=29/43) of pts provided a T2 sample (~71 days), and 41.9% (n=18/43) of pts provided a T3 sample (~117 days). mPFS of pts with CPS ≥10, 1-10, <1 was 8.5, 7.0, 2.0 months, respectively (≥10 CPS vs <10 CPS HR=1.5, p=0.9180), and mOS was 12.1, 9.0, 18.9 months (≥10 CPS vs <10 CPS HR=0.6, p=0.9981). No correlations were identified between CPS and T0 TMFC PD-L1 (p=0.6109). Further, pts with high TMFC PD-L1 at T2 (HR=3.1, p=0.0475) had significantly better PFS, while T1 (HR=1.8, p=0.1843) and T3 (HR=4.0, p=0.0686) trended toward better PFS, but not for OS. Pts with high TMFC PD-L1 at any time point had significantly improved PFS (HR=2.8, 95% CI=1.4-5.5, p=0.0052), but not OS, compared to pts with consistently low TMFC PD-L1. In this pilot study, tumor PD-L1 CPS was not found to be correlated with clinical outcomes. However, high TMFC PD-L1 expression at any timepoint correlated with improved PFS, suggesting that monitoring PD-L1 in TMFCs may serve as a real-time biomarker to better indicate ICI response. Further studies into the role of TMFC PD-L1 in predicting therapeutic response are ongoing.