Genomics of NSCLC patients both affirm PD-L1 expression and predict their clinical responses to anti-PD-1 immunotherapy

Abstract

Background: Programmed Death Ligand 1 (PD-L1) is a co-stimulatory and immune checkpoint protein. PD-L1 expression in non-small cell lung cancers (NSCLC) is a hallmark of adaptive resistance and its expression is often used to predict the outcome of Programmed Death 1 (PD-1) and PD-L1 immunotherapy treatments. However, clinical benefits do not occur in all patients and new approaches are needed to assist in selecting patients for PD-1 or PD-L1 immunotherapies. Here, we hypothesized that patient tumor cell genomics influenced cell signaling and expression of PD-L1, chemokines, and immunosuppressive molecules and these profiles could be used to predict patient clinical responses.

Methods: We used a recent dataset from NSCLC patients treated with pembrolizumab. Deleterious gene mutational profiles in patient exomes were identified and annotated into a cancer network to create NSCLC patient-specific predictive computational simulation models. Validation checks were performed on the cancer network, simulation model predictions, and PD-1 match rates between patient-specific predicted and clinical responses.

Results: Expression profiles of these 24 chemokines and immunosuppressive molecules were used to identify patients who would or would not respond to PD-1 immunotherapy. PD-L1 expression alone was not sufficient to predict which patients would or would not respond to PD-1 immunotherapy. Adding chemokine and immunosuppressive molecule expression profiles allowed patient models to achieve a greater than 85.0% predictive correlation among predicted and reported patient clinical responses.

Conclusions: Our results suggested that chemokine and immunosuppressive molecule expression profiles can be used to accurately predict clinical responses thus differentiating among patients who would and would not benefit from PD-1 or PD-L1 immunotherapies.

Trial registration: ClinicalTrials.gov NCT01295827.

Keywords: Computational modeling; Immunotherapy; NSCLC; PD-1; PD-L1.

Conflict of interest statement

Ethics approval and consent to participate

All patients had stage IV non-small cell lung cancer (NSCLC) and were treated at Memorial Sloan Kettering Cancer Center (n = 29) or the University of California at Los Angeles (n = 5) on protocol NCT01295827. All patients had consented to the Memorial Sloan Kettering Cancer Center Institutional Review Board-approved protocols permitting tissue collection and sequencing.

All patient related research was Memorial Sloan Kettering Cancer Center Institutional Review Board-approved and treated under protocol NCT0129827. Written informed consent was obtained from all patients.

Consent for publication

Not applicable.

Competing interests

KAB has had a Cooperative Research and Development Agreement with Cellworks Group Inc., San Jose, CA. TA and SV work for Cellworks Group Inc., San Jose, California. DP, NKS, and UM work for Cellworks Research India Ltd., Whitefield, Bangalore, India. TAC is a co-founder of Gritstone Oncology and holds equity in the company. All other authors declare no competing financial interests in the findings of this study or with Cellworks Group Inc., San Jose, CA or Cellworks Research India Ltd., Whitefield, Bangalore, India.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Figures

Fig. 1

The schema for creating predictive computational simulation models to predict molecule responses and identify patients that would respond or not respond to PD-1 immunotherapy treatment using patient SA97V5 as a model example. Exome information from patient SA97V5 (a) contained 1192 total mutations with 36 deleterious mutations. This profile (b) was converted from a mutational profile to a computational format and annotated into the computational workflow to convert (c) a nontransformed model in the cancer network into (d) a patient SA97V5-specific simulation model. The patient SA97V5-specific simulation model was used to predict PD-L1 expression (e.g., 67.0% with respect to control), dendritic cell (DC) infiltration index (e.g., 23.8% with respect to control); and an immunosuppressive molecule expression profile (e.g., range − 1.9% to 56.5% with respect to controls) (e). Predicted expression responses were all used (f) to sort patients into groups that would respond or not respond to PD-1 immunotherapy treatment. SA97V5 was identified as a patient who would respond to PD-1 immunotherapy treatment. Numerous validation checks (g) occurred on the cancer network, the simulation model predictions, and the PD-1 match rates between the predicted responses and the patient clinical responses

Fig. 2

A decision tree was used to identify PD-1 drug responder status. At step 1, 9 patients with PD-L1 expression below 29.0% were identified as PD-1 drug non-responders. The remaining 16 patients (including patient SA97V5) with PD-L1 expression equal to or greater than 29.0% proceeded to step 2. At Steps 2a and 2b, 2 patients with dendritic cell infiltration index values below 20.0% were identified as non-responders and 2 patients with index values greater than 60.0% were identified as PD-1 drug responders. Twelve patients with index values greater than 20.0% (including patient SA97V5), but less than 60.0% proceeded to step 3. At Step 3, 4 patients with immunosuppressive molecule (ISM) values higher than that of their PD-L1 expression with a margin of greater than 5.0%, were identified as non-responders (Step 3a) and 8 patient-specific models with values lower than that of their PD-L1 expression with a margin of greater than 5.0% were identified as responders (Step 3b, including patient SA97V5). Mismatch patients GI7AGZ, 2FCOH7, F3FK2W were not listed

Fig. 3

Patient-specific simulation models were used to predict the expression of PD-L1 (a) and at step 1 of the decision tree, 9 patients (black bars) with predicted PD-L1 expression below 29.0% (bold line) were identified as PD-1 drug non-responders. Patient SA97V5 had a predicted PD-L1 expression of 67.0%. Patient-specific simulation models were used to predict the expression of chemokines used to create a dendritic cell (DC) infiltration index (b). At step 2a of the decision tree, 2 patients (black bars) with index values greater that 60.0% (bold line) were identified as PD-1 drug responders and 2 patients (black bars) with index values less than 20.0% (black line) were identified as PD-1 drug non-responders. Patient SA97V5 had a predicted DC infiltration index of 23.9%

Fig. 4

Patient-specific simulation models were used to predict the expression of 14 immunosuppressive molecules. At step 3 of the decision tree, patients with immunosuppressive molecule predictions higher than that of PD-L1 with a margin of greater than 5.0% (bold line), were considered to be non-responders and patient-specific models with predictions lower than that of PD-L1 with a margin of greater than 5.0% were considered to be responders. Eight remaining patients were identified as responders and 4 remaining patients were identified as non-responders. Patient SA97V5 (a) had all 14 molecules below the threshold and was identified as a PD-1 drug responder. Patient QIA43T (b) had 2 molecules above the threshold and was identified as a PD-1 drug non-responder

Fig. 5

The expression of PD-L1 was influenced via a number of signaling pathways. Activating signals were processed via the ERK signaling pathway (via EGFR; B-Raf proto-oncogene, serine/threonine kinase, BRAF-V600E; mitogen-activated protein kinase kinase 1/2, MEK1/2; mitogen-activated protein kinase kinase 1, MAP2K1; MAP2K2; ERK1/2; mitogen-activated protein kinase 3, MAPK3; mitogen-activated protein kinase 1, MAPK1; and Jun proto-oncogene, c-Jun). Activating signals were processed via the EGFR signaling pathway (via neuroblastoma RAS viral oncogene homolog, NRAS; phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha, PIK3CA; V-akt murine thymoma viral oncogene homolog, AKT; mechanistic target of rapamycin, MTOR; and STAT3). Also, activating signals were processed via the interferon gamma (IFNG) pathway (via IFNG; interferon gamma receptor 1, IFNGR1; signal transducer and activator of transcription 1, STAT1; and interferon regulatory factor 1, IRF1). Pathway signals converge to activation factors Activator protein 1 (AP1), STAT1, STAT3, and IRF1 leading to transcription of PD-L1 genes. Common pathways were utilized among a number patient-specific simulation models. Patient C9TGAJ (KRAS mutation), patient RDD2UW (KRAS mutation), patient M9GYO4 (MAP2K2 mutation), and patient DFZLO2 (MAP3K1 mutation) involved the ERK activation pathway. Patient P90A0O (BRAF1, TP53 mutations) and patient L8MTGU (KRAS, TP53 mutations) involved the ERK activation and apoptotic pathways