Poster T3 - The Use of TF-Ag-a to Select Tumor-Derived Exosomes and Enhance the Predictive Performance of Exosomal PD-L1 in Predicting Response to Immune Checkpoint Inhibitors

Introduction: Immunotherapy, particularly immune checkpoint inhibitors (ICI) targeting PD-1 and PD-L1, has transformed the treatment landscape for advanced cancers such as lung cancer and melanoma. Tumor-derived exosomal PD-L1 has been reported as a promising biomarker in predicting response to ICI therapy. However, in addition to tumor cells, many other cells such as immune cells also release exosomes expressing PD-L1, which may interfere the detection of tumor-derived exosomal PD-L1 and lead to decreased predictive accuracy.

To address this challenge, we developed a surface plasmon resonance (SPR)-based assay which first used Thomsen-Friedenreich glycoantigen (TF-Ag-a, Galβ1-3GalNAc alpha) as the biomarker to selectively capture tumor-derived exosomes (TEX) from all other exosomes, and then analyze TEX PD-L1 expression (Fig. 1a). TF-Ag-α is a cancer-associated carbohydrate antigen arising from aberrant glycosylation found in malignant cells, such as those of lung, breast, prostate, and colon cancer. TF-Ag-α typically has tumor-restricted expression, making it a potent biomarker for TEX capture. In this study, we selected lung cancer as the disease model. We measured TEX PD-L1 levels in sera from advanced lung cancer patients who received pembrolizumab treatment. Our findings showed that TEX PD-L1 achieved 94% accuracy in distinguishing responders from non-responders. TEX PD-L1 significantly outperformed total exosomal PD-L1 in predicting response to ICI therapy.

Materials and

Methods: The levels of exosomal TF-Ag-α and TEX PD-L1 were measured by a SPR-based assay (Fig. 1a). Briefly, the surface of a 49nm Au-coated glass slide was modified with PEG and JAA-F11 antibodies, which had high specificity to TF-Ag-α. When 10µL of serum was applied, the capture of exosomes expressing TF-Ag-α (i.e. TF-Ag-α+ exosomes) via JAA-F11 antibodies changed the local reflective index, enabling the optical detection of exosomal TF-Ag-α levels via SPR. Subsequently, APC dye-labeled anti-PD-L1 antibodies were introduced. After 1 h incubation, excess APC-anti-PD-L1 antibodies were washed off. The fluorescence signals were measured to quantify TEX PD-L1 levels.

Serum samples were collected from lung cancer patients treated with pembrolizumab at both pre-treatment baseline and the end of the first treatment cycle (3-8 weeks post treatment). The predictive performance of TEX PD-L1 was first evaluated in a training cohort of lung cancer patients (15 responders and 14 non-responders) and then validated in an independent test cohort (25 responders and 37 non-responders). The predictive performance metrics (including sensitivity, specificity, and AUC) of exosomal TF-Ag-α and TEX PD-L1 were determined through receiver operating characteristic (ROC) analysis.

Results, Conclusions, and Discussions: In the training cohort, no significant difference in the levels of exosomal TF-Ag-α and TEX PD-L1 was observed between responders (n=15) and non-responders (n=14) at pre-treatment baseline. However, at the end of treatment cycle 1, significantly lower levels of exosomal TF-Ag-α and higher levels of TEX PD-L1 were observed in responders than non-responders. Results from ROC analysis showed that exosomal TF-Ag-α and TEX PD-L1 distinguished responders from non-responders with a sensitivity of 0.86, specificity of 0.80, and AUC of 0.88.

In the independent test cohort, at pretreatment baseline, both responders (n=25) and non-responders (n=37) had comparable exosomal TF-Ag-α levels. However, responders showed significantly lower levels of TEX PD-L1 compared with non-responders. At the end of treatment cycle 1, similar as pretreatment baseline, both responders and non-responders showed comparable exosomal TF-Ag-α levels. However, responders showed increased TEX PD-L1 levels in contrast to non-responders (Fig. 1b). The fold change (end of cycle 1/pretreatment baseline) of exosomal TF-Ag-α and TEX PD-L1 successfully differentiated responders from non-responders with a sensitivity of 0.97, specificity of 0.84, and AUC of 0.94 (Fig. 1c), which notably surpassed tumor PD-L1, predicting only 28.9% of cases.

To confirm that TEX was more accurate in predicting ICI response, we compared the predictive performance between TEX PD-L1 and total exosomal PD-L1 in selected patients (17 responders and 36 non-responders). The fold change (end of cycle 1/pretreatment baseline) of TEX PD-L1 successfully differentiated responders from non-responders with a sensitivity of 0.94, specificity of 0.82, and AUC of 0.94, which was superior to the fold change of total exosomal PD-L1 (sensitivity of 0.92, specificity 0.77, and AUC 0.87).

These findings demonstrated that using TF-Ag-α to select TEX and using TEX PD-L1 can further improve the predictive accuracy of ICI treatment response in lung cancer patients. In the future, we will further validate the predictive performance of TEX PD-L1 in a large cohort of patients. We will also evaluate other tumor associated biomarkers in the selection of TEX to further improve the predictive accuracy.

Acknowledgements (Optional):