Virtual Library
Start Your Search
Liliang Xia
Author of
-
+
P33 - Pathology - Immunotherapy Biomarker (ID 101)
- Event: WCLC 2020
- Type: Posters
- Track: Pathology, Molecular Pathology and Diagnostic Biomarkers
- Presentations: 1
- Moderators:
- Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
-
+
P33.07 - Peripheral CD4+ T cell Signatures in Predicting Response to Anti-PD-1 Monotherapy for Chinese Advanced Non-Small Cell Lung Cancer Patients (ID 1981)
00:00 - 00:00 | Author(s): Liliang Xia
- Abstract
Introduction
Immune checkpoint inhibitors (ICIs) greatly improve clinical outcomes of advanced non-small cell lung cancer (NSCLC). However, limitation of benefit population makes it urgent to screen predictive biomarkers for stratify the patients. In this study we have investigated the peripheral CD4+ T cell signatures and their association with the responses to ICIs treatment.
Methods
52 advanced NSCLC patients were recruited in a registered anti-PD-1 antibody clinical trial and evaluated every 8 weeks according to RECIST v1.1 guidelines. Patients were subgouped according to disease progression at 8 weeks. Blood was collected before the treatment and at the evaluation time point after 8 weeks. Multi-color flow cytometry was performed to define CD4+ T cell signatures. Its relevance to better response to anti-PD-1 monotherapy was evaluated by Wilcoxon test and Kaplan-Meier analysis.
Results
PD-1 distribution in peripheral immune subsets from NSCLC patients (n=52) revealed that CD4+ T cells was of the highest percentage among PD-1+ immune cells (39.4%) whereas CD8+ T cells (29.6%), NK cells (15.6%) and B cells (14.4%) were less. 76.7% of PD-1+CD4+ T cells were memory subsets. When subgrouped 52 patients into responder (R, n=28) and non-responder (NR, n=24) groups according to disease progression after the evaluation at 8 weeks, it was found that average percentage of naïve CD4+ T(Tn) cells at the baseline was lower in R than in NR while memory CD4+ T (Tm) cells was vise verse. The percentages of cytokine-secreting CD4+ Tn cells, including IFN-γ, TNF-α, IL-10 and IL-17A, were significantly higher in R than those in NR. Patients with higher percentages of IFN-γ or IL-17A-producing CD4+ Tn cells possessed a longer PFS. In CD4+ Tm subset, the percentages of CTLA-4+ and Tim3+ CD4+ Tm subsets were higher in R than those in NR while either IFN-γ and TNF-α only or double secreting CD4+ Tm cells were higher in R as well. However, only CTLA-4+ and IFN-γ-producing CD4+ Tm cells were associated with a longer PFS. What's more, while the percentage of CTLA-4+CD4+ Tm cells maintained from the baseline to 8 weeks after the treatment in R, it increased in NR leading to the increase in the ratio of 8 weeks to baseline percentages. Logistic regression analysis revealed that baseline IFN-γ-producing CD4+Tn and CD4+ Tm cells were the most important signatures in the prediction of response to anti-PD-1 therapy for advanced NSCLC patients with an AUC values of 0.758 according to ROC curves.
Conclusion
Our study elucidates that peripheral CD4+ Tn and Tm cell subsets with distinct phenotypes and functions at baseline show different distribution between R and NR. Association analysis with PFS narrows down CD4+ T cell signatures to higher IFN-γ-producing CD4+Tn and CD4+ Tm percentages at baseline with good response to anti-PD-1 therapy in Chinese advanced NSCLC patients. Considering IFN-γ-producing CD4+Tn with functional property and CD4+ Tm pool with rapid responses to the stimuli, these functional signatures at the baseline might be more prone to respond to ant-PD-1 treatment with a better prognosis. Further validation will warrant in the prospective study with larger population.
-
+
P42 - Screening and Early Detection - Risk Modelling and Artificial Intelligence (ID 177)
- Event: WCLC 2020
- Type: Posters
- Track: Screening and Early Detection
- Presentations: 1
- Moderators:
- Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
-
+
P42.03 - A Dynamic Deep Learning Approach to Predict Clinical Outcomes of Patients with Advanced Non-Small Cell Lung Cancer under Nivolumab Monotherapy (ID 1910)
00:00 - 00:00 | Author(s): Liliang Xia
- Abstract
Introduction
Selecting patients with advanced non-small cell lung cancer (NSCLC) who will likely have survival benefit from immune checkpoint therapy seems not easy. In the study, we developed and evaluated a dynamic machine deep learning approach to predict clinical outcomes of advanced NSCLC patients treated with nivolumab monotherapy.
Methods
99 patients treated with nivolumab were included in this retrospective study. 107 radiomic features were extracted from CT imaging for each patient. A deep learning model incorporating radiomic features, laboratory examination data (e.g.blood cell counts and lactate dehydrogenase levels) 90 days prior to the time of efficacy assessment, and baseline clinical information (e. g.TNM,PS score) was developed to robustly differentiate nivolumab responders (R) from non-responders (NR) in a training set of 60 patients. Prediction performance was then tested in a validation set of 39 patients. A simple temporal attention (SimTA) model was designed to process asynchronous imaging time-series and laboratory data (Figure 1).
Results
The deep leaning-based predicting model achieves significantly predictive performance (area under the curve [AUC] is 0.79 [95% CI:0.63-0.94], p=0.00). This model stratified the patients into high- and low-risk NR groups according to a default threshold. Patients in low-risk group were associated with improved PFS (7.00 ±1.03 [95% CI: 4.98-9.02] months vs. 1.30±0.17 [95% CI: 0.97-1.63] months; Log-rank p=0.00 ) and OS (28.00 ±0.46 [95% CI: 27.10-28.90] months vs. 5.70 ±1.80 [95% CI: 2.17-9.23] months; Log-rank p=0.00) compared with high-risk group (Figure 2). Finally, an exploratory analysis of 11 patients with stable disease (after first efficacy assessment by RECIST 1.1) in the validation set showed that low-risk patients had superior OS than high-risk patients (28.6±11.98 [95% CI: 5.13-52.07] months vs. 8.80 months, Log-rank p=0.00).
Conclusion
With easily available radiographic, laboratory and clinical information, SimTA-based dynamic deep learning provides a promising method of predicting treatment response and long-term survival in patients with advanced NSCLC under nivolumab monotherapy.
-
+
P72 - Tumor Biology and Systems Biology - Basic and Translational Science - Tumor Microenvironment (ID 211)
- Event: WCLC 2020
- Type: Posters
- Track: Tumor Biology and Systems Biology - Basic and Translational Science
- Presentations: 1
- Moderators:
- Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
-
+
P72.06 - NLRP4-Mediated Type I Interferon Response Benefits Immune Checkpoint Therapy Through Redirecting CD8+ T Cell Distribution in TME (ID 1706)
00:00 - 00:00 | Presenting Author(s): Liliang Xia
- Abstract
Introduction
The application of immune checkpoint inhibitors (ICIs) in the treatment for advanced non-small cell lung cancer (NSCLC) achieves remarkable accomplishment in the last several years with the 5-year survival rate in advanced NSCLC patients increasing from 5% to 17%, including anti-programmed cell death protein 1 (PD-1) or programmed death-ligand 1(PD-L1) antibodies. Despite of the progress, only 20% of patients benefit from monotherapy of ICIs and fewer with durable responses. The occurrence of resistance to ICIs becomes a major bottleneck of their clinical applications. Thus, the main challenges in cancer immunotherapy targeting ICs is how to develop new therapies to overcome the resistance and expand the scope of benefit population as well as to screen with great precision the population who can benefit from it. Defining the inherent pattern of tumor-intrinsic immune property in tumor microenvironment becomes an opportunity to find new strategies to overcome ICI resistance.
Methods
We have performed the whole exome sequencing (WES) of tumor biopsies and paired peripheral blood from advance NSCLC patients with different responses to anti-PD-1 antibody therapy before the treatment. Bioinformatic analysis was performed to screen somatic gene mutations in tumors. A nod-like receptor (NLR) family member NLR Family Pyrin Domain Containing (NRPL4) was defined. Its roles in regulating tumor growth as well as the coordination with anti-PD-L1 antibody immunotherapy were investigated both in vitro and in vivo of mouse models. The mechanisms of NLRP4-mediated tumor-intrinsic modulation on tumor microenvironment were extensively investigated.
Results
There existed NLRP4 mutations in PD-1 blockade responders whereas no mutation was detectable in non-responders. Knockdown of NLRP4 in mouse lewis lung cancer (LLC) cell line enhanced IFN-α/β production via cGAS-Sting-IRF3/IRF7 axis. While CD8+ T cells were enriched more apparently in the tumors with the deletion of NLRP4 in LLC, in vitro transwell assay also demonstrated that knockdown of NLRP4 in LLC triggered the migration of CD8+ T cells in a more extent manner. With the slight increase of PD-L1 expression in NLRP4 deficient LLC cells, knockdown of NLRP4 delayed tumor growth and synergized the efficacy with PD-L1 therapy largely relying on the increase of type I interferon in tumors. This was consistent with clinical observation that a high level of periphery IFN-α was associated with a longer progression free survival of NSCLC patients undergoing anti-PD-1 treatment and more CD8+ T cells was surrounding tumor cells to the curative effect of immunotherapy.
Conclusion
Our study thus strongly implies that NLRP4 mutations in lung cancer patients might drive enhanced type I interferon responses, which in turn redirects the localization of CD8+ T cells in tumor microenvironment and is beneficial for anti-PD-1 treatment. NLRP4-mediated tumor-intrinsic type I interferon responses thus not only orchestrate ICIs efficacy to overcome ICI resistance in patients lacking of responses to the treatment, but also become favorable and predictive biomarkers to guide the stratification of NSCLC patients for ICIs clinical applications.
