Session Item

Despite the growing interest in combining radiotherapy (RT) with immune checkpoint blockers (ICB), there is a lack of understanding on the effect of RT itself on the human tumour-immune microenvironment (TIME) and thus the anti-tumour immune response. The PRADA (NCT02771938) and Neo-RT (NCT03818100) trials, which tested feasibility and safety of neoadjuvant RT (NRT) in breast cancer patients, represent a unique opportunity to study novel in vivo longitudinal radiobiological data from the treatment-naïve TIME. We characterised NRT-induced changes to stromal tumour-infiltrating lymphocytes (sTILs) and peripheral blood lymphocytes in these cohorts.

PRADA cases (n=30) received pre-operative locoregional RT (40Gy/15F) following neoadjuvant chemotherapy (NACT). Contemporaneous case-matched controls for PRADA (n=30) received NACT alone. Neo-RT (n=13) received pre-operative locoregional RT (40Gy/15F, with simultaneous integrated boost of 48Gy/15F to primary tumour where possible) followed by 20 weeks of pre-operative endocrine therapy. FFPE samples were obtained from primary tumour at diagnosis (T0), after one fraction of NRT (T1), after final fraction of NRT (T2), and at surgery (T3). H&E-stained sections were manually scored by a breast pathologist blinded to clinical context, following the International TILs Working Group guidelines. Digitised H&E sections were also scored using artificial intelligence (AI) methods for comparison. Absolute lymphocyte count (ALC) from full blood count data at relevant timepoints was evaluated for a subset of patients from PRADA.

PRADA cases experienced a significant and uniform decrease in sTILs between T0 and subsequent timepoints, with no recovery by T3 (Fig.1A). This was not seen in PRADA controls following NACT alone. sTILs also showed a trend towards decreasing infiltration in Neo-RT. Both NACT and NRT contributed to significant decreases in ALC (Fig.1B). High sTILs at T0 was associated with pathological complete response (pCR) in PRADA cases only, but a relationship between sTILs at T3 and relapse-free survival was seen in both PRADA cases and controls (Fig.2A,B). AI-TILs analysis using a novel approach combining cell classification with the SuperHistopath pipeline showed superior concordance with manual scores when compared with other AI methods. A model using combined manual-AI scores had greater AUC for prediction of pCR compared to models using either parameter alone.

Our data suggest a locally lymphodepletive effect of NRT on breast TIME with NRT dose-volumes and sequencing schedules used in these trials. However, the correlations between outcome and sTILs at both baseline and time of surgery in patients receiving NRT suggest a complex underlying relationship between RT and the immune response. Ongoing work on these samples including multiplex immunofluorescence and TCR-seq will elucidate this further, and ultimately inform the rational design of future RT-ICB combination trials.