Azacitidine (AZA) with Nivolumab (Nivo), and AZA with Nivo + Ipilimumab (Ipi) in Relapsed/Refractory (R/R) Acute Myeloid Leukemia: Clinical and Immune Biomarkers of Response

Background: Blockade of PD-1/PD-L1 pathways enhances anti-leukemic responses in pre-clinical studies. PD1 inhibition alone had limited clinical activity in AML. CLTA4 inhibition demonstrated encouraging single-agent CR’s in postASCT patients (pts), especially for extra-medullary (EMD) relapses (Davis M et al, NEJM 2016). AZA+Nivo up-regulated CTLA4 on bone marrow (BM) CD8 cells in both responders (R) and non-responders (NR) on treatment (Daver N, et al, Cancer Discovery 2019), suggesting a triplet of AZA+Nivo+Ipi may abrogate PD-1 mediated resistance.

Methods: Pts were eligible for the AZA+Nivo (cohort 1, n=59) if they had R/R AML, ECOG ≤ 2, and adequate organ function. This cohort has completed enrollment.
A cohort of AZA+Nivo+Ipi was opened (cohort 2), with the same eligibility criteria. Ipi 1mg/kg Q6 weeks was added to the established AZA+Nivo schedule and found to be safe in a lead-in dose cohort.

Results: Cohort 1 (59 R/R AML) pts were treated with Aza+Nivo (Daver et al., Cancer Discovery 2019). CR/CRi and OS were superior to contemporary historic HMA-based clinical trial controls at MDACC (Table 1). Pts with low pretherapy BM blasts (<20% BM blasts) and early salvage (salvage 1) had encouraging median (med) OS 11 months (mos), indicating progressive T-cell exhaustion with multiple relapses. Importantly, on 40-parametric CYTOF assessment, Rs to AZA+Nivo had a higher frequency of pre-therapy BMA CD3+ and CD8+ cells compared with NRs (optimal CD3+ cutoff 13.2%: ORR 56% versus 23%), suggesting pre-existing BM T-cell infiltration may be a pre-requisite of response to PD1 based therapies, as PD1 inhibitors have limited ability to mobilize peripheral T-cells (Fig 1A). The pre-therapy BM CD4 polyfunctional strength index (PSI) defined as percentage of polyfunctional cells in the sample, multiplied by the intensities of the secreted cytokines, assessed by single cell cytokine analysis was dramatically different between Rs and NRs (P=0.03). Single-cell RNA-seq on 113,394 BM cells collected longitudinally from 8 pts (3R, 2 stable disease (SD), 3NR) demonstrated that deletion 7/7q, LSC signature enrichment, and activated metabolic/oxidative pathways, were associated with resistance to AZA+Nivo (Abbas H et al, ASH 2020). Importantly, AZA+Nivo induced novel and expanded T cell clonotypes, almost exclusively in Rs.

Cohort 2 (36 R/R AML) pts were treated with Aza+Nivo+Ipi, with med age 67 years (25-83), secondary AML (50%), ELN unfavorable cytogenetics (67%), TP53 (36%), med salvage 2 (range, 1-3), and prior HMA based therapies in 67%. All 36 pts are evaluable. Per ELN 2017, CR/CRi was noted in 7 (19%) and PR in 1 (3%). Five (14%) pts had durable stable disease (SD) (defined as absence of CR, CRi, PR; with SD on treatment for ≥6 months), and 23 (64%) were NRs (Table 1). Interestingly, 4 pts with EMD were enrolled and 3 had CR/PR [med DOR 8 (5-13) mos]. The 4- and 8-week mortalities were 0 and 6%, respectively. Grade 3/4 immune toxicities noted in 8 pts (19%), including rash, pneumonitis, colitis, pyrexia. One pt required ICU stay, but no deaths attributed to immune toxicity. Other grade ≥2 toxicities were as expected for R/R AML population and were mostly infections/febrile neutropenia. Converse to AZA+Nivo, on CYTOF, R did not have a higher frequency of pre-therapy BM CD8+ infiltration, but did have progressive BM CD8+ infiltration on therapy, compared with NRs, demonstrating that Ipi (unlike Nivo) may be able to mobilize peripheral T-cells to the BM (Fig 1B). Expansion of a cluster of antigen experienced CD8+ T cells was associated with response (Fig 1B).

In all salvage the med OS with Aza+Ipi+Nivo versus Aza+Nivo versus contemporary HMA-controls in R/R AML, were 7.6, 5.9, and 4.6 mos, respectively (P=0.01) (Fig 1C). The 1-year OS in R/R AML pts with AZA+Nivo+Ipi was 25%. The med OS with Aza+Ipi+Nivo was comparable to med OS 6-8 mos reported with HMA+VEN salvage in numerous studies.

Conclusion: The OS with Aza+Nivo+Ipi was modestly improved over AZA+Nivo and HMA-controls in R/R AML. Single-cell cytokine profiling and single-cell RNA-seq from Aza+Nivo showed striking pre- and on-treatment differences among R and NR not only in T-cell fitness and clonality, but also in the tumor microenvironment. Ipi may be uniquely able to mobilize peripheral T-cell to BM and EMD. This underappreciated immune diversity suggests a critical need for biomarker-based trials (as we are doing with molecular therapies) for immunotherapies in AML.