Expression patterns of immune checkpoints in acute myeloid leukemia

To the Editor,

Immune checkpoint (IC) blockade by inhibitors of the programmed cell death 1 (PD-1) and PD-1 ligand 1 (PD-L1) has significantly improved clinical outcome for a variety of solid tumors [1, 2], while little is known about the role of ICs in leukemia [3]. Previous reports have shown that higher numbers of PD-1 + T cells are related to poor outcome for patients with acute myeloid leukemia (AML) [3]. Clinical trials using PD-1 inhibitors are ongoing to treat patients with a high risk for AML relapse [4]. However, the response rate varies widely, ranging from 22 to 72% [4], which may be due to heterogeneity in the IC expression level as well as distinct dominant IC expression patterns in different AML cases [5]. Therefore, it is worth studying the expression patterns of ICs in AML. In this study, we first explored the prognostic value of ICs in AML patients through analyzing RNA-seq and mutation data from the Cancer Genome Atlas (TCGA) database [6] and further validated the results by quantitative real-time PCR analysis of AML bone marrow (BM) samples from our clinical center.

A total of 176 de novo AML patients from the TCGA database and 62 AML BM samples were used for overall survival (OS) analysis and validation. Higher expression of PD-1, PD-L1, and PD-L2 correlated with poor OS in the TCGA database analysis (3-year OS 23% vs 38%, 19% vs 46%, and 15% vs 40%, respectively, P < 0.05). This result was confirmed in the validation group (3-year OS 40% vs 68%, 22% vs 64%, and 42% vs 68%, respectively, P < 0.05, Fig. 1a, b). We further analyzed the expression patterns of PD-1, PD-L1, and PD-L2 with other important ICs [7‐9]. Subsequently, with Pearson’s correlation analysis, we found that the expression of PD-1, PD-L1, or PD-L2 was positively associated with the expression of cytotoxic T-lymphocyte associated protein 4 (CTLA-4) (r = 0.259, P < 0.001; r = 0.435, P < 0.001; r = 0.269, P < 0.001, respectively) and lymphocyte activation gene-3 (LAG-3) (r = 0.275, P < 0.001; r = 0.276, P < 0.001; r = 0.160, P = 0.033, respectively) in the TCGA group (Fig. 1c). This concomitant expression pattern was again confirmed in the validation group (Fig. 1e), showing the possibility of concomitant expression of PD-1, PD-L1, or PD-L2 with CTLA-4 (r = 0.373, P = 0.003; r = 0.998, P < 0.001; r = 0.998, P < 0.001, respectively) and LAG3 (r = 0.372, P = 0.003; r = 0.994, P < 0.001; r = 0.994, P < 0.001, respectively). AML patients with high expression of CTLA-4 and LAG-3 were found to have poor OS (3-year OS 9% vs 36% and 13% vs 40% respectively) (Fig. 1d). This result was again confirmed in the validation group (Fig. 1f) (3-year OS: CTLA-4 34% vs 66%, LAG-3 33% vs 70%).

Combination of IC inhibitors (ICIs) has the potential to improve responses [4, 10]. We analyzed expression patterns of ICs and found that pairwise combinations of PD-1, PD-L1, and PD-L2 and CTLA-4 and LAG-3 correlated with poor OS in AML patients (P < 0.05, Figure S1). Furthermore, among AML patients with high expressions of PD-1 or PD-L2, concomitant high expression of CTLA-4 correlated with poor OS in both the TCGA database (3-year OS: PD-1^highCTLA-4^low vs PD-1^highCTLA-4^high 30% vs 0%, 20% vs 0%) and validation group (3-year OS: PD-L2^highCTLA-4^low vs PD-L2^highCTLA-4^high 57% vs 31%, 57% vs 33%) (P < 0.05, Fig. 2a, b). AML with PD-L1^highCTLA-4^high correlated with poor OS in the TCGA dataset (3-year OS 24% vs 0%, P < 0.001); however, OS was not significantly different in the validation group (3-year OS 33% vs 20%, P = 0.353, Fig. 2a, b). In addition, high expression of LAG-3 with PD-1^high, PD-L1^high, or PD-L2^high failed to correlate with OS in the TCGA and validation groups (Figures S2A - B).

To obtain the effects of PD-1, PD-L1, and PD-L2 on the prognosis of AML patients with genetic mutations, we analyzed OS of the top ten AML patients with a recurrent mutation (Figures S3), including FLT3^mut, RUNX1^mut, or TET2^mut [11‐13]. Interestingly, in comparison with AML patients without such mutations, high co-expressions of PD-1/PD-L1 (P = 0.029), PD-1/PD-L1/PD-L2 (P = 0.003), and PD-1/LAG-3 (P < 0.001) were found to be associated with poor OS in AML patients with FLT3^mut, RUNX1^mut, or TET2^mut (1-year OS 0% vs 58% vs 49%, 0% vs 56% vs 100%, and 0% vs100% vs 63%, respectively) (Fig. 2c).

To the best of our knowledge, we for the first time described that high co-expressions of PD-1/CTLA-4 and PD-L2/CTLA-4 correlated with poor OS of AML patients. Moreover, high co-expressions of PD-1/PD-L1, PD-1/PD-L1/PD-L2, and PD-1/LAG-3 were associated with poor OS of AML patients with FLT3^mut, RUNX1^mut, or TET2^mut, respectively (Fig. 2d). These co-expression patterns might be potential immune biomarkers for designing novel AML therapy.