Blood immune cells as potential biomarkers predicting relapse-free survival of stage III/IV resected melanoma patients treated with peptide-based vaccination and interferon-alpha

Abstract

Introduction: Despite the recent approval of several therapies in the adjuvant setting of melanoma, tumor relapse still occurs in a significant number of completely resected stage III-IV patients. In this context, the use of cancer vaccines is still relevant and may increase the response to immune checkpoint inhibitors. We previously demonstrated safety, immunogenicity and preliminary evidence of clinical efficacy in stage III/IV resected melanoma patients subjected to a combination therapy based on peptide vaccination together with intermittent low-dose interferon-α2b, with or without dacarbazine preconditioning (https://www.clinicaltrialsregister.eu/ctr-search/search, identifier: 2008-008211-26). In this setting, we then focused on pre-treatment patient immune status to highlight possible factors associated with clinical outcome.

Methods: Multiparametric flow cytometry was used to identify baseline immune profiles in patients' peripheral blood mononuclear cells and correlation with the patient clinical outcome. Receiver operating characteristic curve, Kaplan-Meier survival and principal component analyses were used to evaluate the predictive power of the identified markers.

Results: We identified 12 different circulating T and NK cell subsets with significant (p ≤ 0.05) differential baseline levels in patients who later relapsed with respect to patients who remained free of disease. All 12 parameters showed a good prognostic accuracy (AUC>0.7, p ≤ 0.05) and 11 of them significantly predicted the relapse-free survival. Remarkably, 3 classifiers also predicted the overall survival. Focusing on immune cell subsets that can be analyzed through simple surface staining, three subsets were identified, namely regulatory T cells, CD56^dimCD16^- NK cells and central memory γδ T cells. Each subset showed an AUC>0.8 and principal component analysis significantly grouped relapsing and non-relapsing patients (p=0.034). These three subsets were used to calculate a combination score that was able to perfectly distinguish relapsing and non-relapsing patients (AUC=1; p=0). Noticeably, patients with a combined score ≥2 demonstrated a strong advantage in both relapse-free (p=0.002) and overall (p=0.011) survival as compared to patients with a score <2.

Discussion: Predictive markers may be used to guide patient selection for personalized therapies and/or improve follow-up strategies. This study provides preliminary evidence on the identification of peripheral blood immune biomarkers potentially capable of predicting the clinical response to combined vaccine-based adjuvant therapies in melanoma.

Keywords: adjuvant therapy; circulating biomarkers; combination therapy; immunotherapy; interferon alpha (IFN-α); melanoma; multiparametric flow cytometry; peptide vaccination.

Figure 1

Pretreatment frequency of the main T cell subsets in relapsing (REL) and non-relapsing (NED) patients. Box plots (showing median, interquartile range, minimum and maximum) representing the pretreatment levels of the main T cell subsets in NED (no evidence of disease) and REL (relapsed) patients. The number of patients (n) is indicated below graphs. (A) Frequencies of CD3⁺ T cells (expressed as percentage of lymphocytes) and of CD4⁺, CD8⁺ and gd T cells (expressed as percentage of CD3⁺ cells). (B) Frequencies of naïve (N), central memory (CM), effector memory (EM) and terminally differentiated (TD) cells within the indicated T cell subset. P-values by Mann-Withney non-parametric U test.

Figure 2

Pretreatment frequency and functionality of immune cell subsets in relapsing (REL) and non-relapsing (NED) patients. Box plots (showing median, interquartile range, minimum and maximum) represent 9 different immune cell subsets with a significantly different baseline frequency in NED and REL patients. The number of patients (n) is indicated below graphs. (A) Frequencies of regulatory T cells (Tregs, CD3⁺CD4⁺CD25^hiCD127-Foxp3⁺), expressed as percentage of CD3⁺ T cells, of terminally differentiated (TD) CD8^hiCD4^low (CD3⁺CD8^hiCD4^lowCD45RA⁺CCR7-), and effector memory (EM) CD8^lowCD4^hi (CD3⁺CD8^lowCD4^hiCD45RA-CCR7-), expressed as percentages of indicated subsets. (B) Polyfunctionality upon short-term in vitro expansion with staphylococcal enterotoxin B (SEB) or Melan-A peptide. Total CD8⁺ T cells producing TNF-α and 3 cytokines simultaneously in SEBstimulated samples (left panels). Melan-A-specific CD8⁺ T cells producing one and zero cytokines after Melan-A recognition (right panels). (C) Natural killer (NK) cell phenotype and functionality. The frequency of CD56^dimCD16^- NK cell subset is expressed as percentage of CD3^- cells (left panel). Levels of CD56^hiCD16⁺ NK cells producing IFN-γ in response to phorbol 12-myristate 13-acetate (PMA)/ionomycin (PMA⁺I), are shown as percentages of CD3^- cells (right panel). P-values by Mann-Withney non-parametric U test.

Figure 3

Predictive accuracy of the identified markers. Receiver operating characteristics (ROC) curves showing sensitivity and specificity of baseline markers that significantly discriminate NED and REL patients. (A) Treg, EM CD3⁺, N CD4⁺, TD CD8^hiCD4^low, EM CD8^lowCD4^hi and CM gd T cells, identified as in Figures 1 and 2 . (B) SEB-stimulated CD8⁺ T cells producing TNF-α and 3 cytokines (left panels) and Melan-A-specific CD8⁺ T cells producing one and zero cytokines upon Melan-A peptide stimulation (right panels). (C) CD56^dimCD16^- and PMA/ionomycin (PMA+I)-stimulated IFN-g⁺CD56^hiCD16⁺ NK cells. The area under the curve (AUC), the asymptotic P-value and the cut-off are indicated for each curve.

Figure 4

Relapse free survival analyses. Kaplan–Meier survival curves showing the relapse-free survival (RFS) of patients stratified according to the markers’ optimal cut-off identified by ROC analyses (see Figure 2 , Supplementary Tables S1 and S2 ). (A) Treg, EM CD3⁺, N CD4⁺, TD CD8^hiCD4^low, EM CD8^lowCD4^hi and CM γδ T cells, identified as in Figures 1 and 2 . (B) SEB-stimulated total CD8⁺ T cells producing TNF-α and 3 cytokines (left panels) and Melan-A-specific CD8⁺ T cells producing one and zero cytokines upon Melan-A peptide stimulation (right panels). (C) CD56^dimCD16^- and PMA/ionomycin (PMA+I)-stimulated IFN-γ⁺CD56^hiCD16⁺ NK cells. Continue and dotted lines correspond to frequency < or ≥ cut-off, as indicated. P-values by log-rank test.

Figure 5

Overall survival analyses. Kaplan–Meier survival curves showing the overall survival (OS) of patients stratified according to the markers’ optimal cut-off identified by ROC analyses (see Figure 2 , Supplementary Tables S1 and S2 ). (A) Treg, EM CD3⁺, N CD4⁺, TD CD8^hiCD4^low, EM CD8^lowCD4^hi and CM γδ T cells, identified as in Figures 1 and 2 . (B) SEB-stimulated total CD8⁺ T cells producing TNF-α and 3 cytokines (left panels) and Melan-A-specific CD8⁺ T cells producing one and zero cytokines upon Melan-A peptide stimulation (right panels). (C) CD56^dimCD16^- and PMA/ionomycin (PMA+I)-stimulated IFN-γ⁺CD56^hiCD16⁺ NK cells. Continue and dotted lines correspond to frequency < or ≥ cut-off, as indicated. P-values by log-rank test.

Figure 6

Predictive accuracy of the combination score and survival curves. (A, B) Multivariate analysis by principal component analysis (PCA) of the three parameters selected for the combination score, regulatory T cells (Treg), CD56^dimCD16^- NK cells and CM γδ T cells. Biplots displaying both patients (points) and parameters (vectors). Confidence ellipses are provided (CI = 95%) for patients after grouping them according to either (A) Clinical outcome (blue circle=NED, yellow triangle=REL) or (B) Arm (blue circle=arm 1, yellow triangle=arm 2). In the component space, separation between individuals grouped by Arm (p-value=0.78) or Outcome (p-value=0.034) was evaluated using Wilks test. (C) ROC curves showing sensitivity and specificity of the combination score. The area under the curve (AUC), the asymptotic P-value, and the optimal cut-off are indicated. (D, E) Kaplan-Meier plots showing (D) the relapse-free survival (RFS) and (E) the overall survival (OS) of patients stratified according to the cut-off. Continue and dotted lines correspond to frequency < or ≥ cut-off, as indicated. The number of patients at risk is indicated below D and E plots according to the patient stratification. P-values by log-rank test.