the Poorer-Risk AML, the Weaker Immunologic Surveillance? --Higher PD-L1 Expression on Non-APL AML Cells Is Associated with Poorer Risk Status According to Cytogenetics and Molecular Abnormailties

Backgroud

The specific recurrent cytogenetic and molecular abnormalities are significantly associated with prognosis and have been used to establish risk stratification which directs the therapeutic strategy in AML. Whether the immune surveillance contributes to the prognosis according to the cytogenetic and molecular risk status is unknown. The checkpoint PD-1/PD-L1 pathway is involved in the tumor surveillance. Down-regulation of PD-1/PD-L1 pathway caused a slower AML progression in mouse models through anergy of CD8+ CTL cells. Clinical trials of PD-1/PD-L1 inhibitor-based regimens in patients with AML are ongoing and seem to have promising results. But few studies investigated clinical significance of PD-L1 expression on leukemic cells in AML patients.

Method

We prospectively observed 120 consecutive adult patients with non-APL AML. PD-L1 mRNA of bone marrow mononuclear cells at diagnosis was detected by RT-PCR and the specific fluorescence indices (SFI) of PD-L1 expression on bone marrow blast cells was determined by flow cytometry. Cytogenetic and molecular genetic risks were assessed according to NCCN AML guideline (2016 version 2). Patients with age >60 years or failure of karyotype analysis were excluded. All remaining 95 patients were treated with "3+7" idarubicine and cytarabine induction regimen. Patients with response of ≥PR were administered the same regimen as the initial chemotherapy, and then consolidated with intermediate-dose cytarabine for four cycles followed by MA, AA, HA and AE regimens subsequently. Patients at poor risk or with unsatisfied controlled minimal residual disease were recommended allogeneic transplantation (allo-HCT). Among the patients with indication of allo-HCT, those who had both donor and desire received allo-HCT after 3 cycles of consolidation in their first CR (CR1).

Result

The diagnoses according to morphology included AML-M0 (n=1), AML-M1 (n=5), AML-M2 (n=55), AML-M4 (n=19), AML-M5 (n=13) and AML-M6 (n=2). Twenty patients were at favorable risk, 43 were at intermediate risk and 32 were at poor risk. We found a correlation between SFIs of PD-L1 and PD-L1 mRNA (r=0.413, P<0.001). After the base 10 logarithm transformation, quantity of PD-L1 mRNA in mononuclear cells of bone marrow were different among all risk groups (P=0.025). Post Hoc multiple comparisons showed PD-L1 mRNA of poor-risk group was higher than that of favorable-risk group (P=0.008). PD-L1 mRNA of the intermediate-risk group was also higher when compared to that of favorable-risk group (P=0.029). SFIs of PD-L1 on leukemic cells was higher in poor-risk group than in either intermediate-risk (P=0.040) or favorable-risk group (P=0.042), although no statistic difference was found among three groups (P=0.057). With ROC curves, we found PD-L1 mRNA could well predict both the early CR defined as the CR achieved after the initial chemotherapy (AUC=0.736, P=0.050) and the final CR defined as CR1 achieved after any cycles of induction chemotherapy (AUC=0.911, P=0.006). SFI of PD-L1 on leukemic cells could only predict the final CR (AUC=0.889, P=0.009). With both the best specificity of 0.833 and sensitivity of 1.000 for predicting final CR and specificity of 0.667 and sensitivity of 0.917 for predicting early CR, the cutoff value of PD-L1 mRNA was identified as 0.0681. SFI of PD-L1 less than 1.852 was a good predictor for achieving final CR with the sensitivity of 0.800 and specificity of 1.000. PD-L1 mRNA≥0.0681 was the only independent poor-risk factor for the early CR (HR=12.697, 95%CI 2.710-59.481, P<0.001).

Conclusion

AML cells with poorer-risk cytogenetic and molecular abnormalities seemed to have a higher PD-L1 expression which may compromise the leukemic immune surveillance by CTL suppression. Higher PD-L1 expression on AML cells was associated with poorer response to IA-based induction regimens.