Identifying Immune Resistance Mechanisms to CD33/CD3 BiTE® Antibody Construct (AMG 330) Mediated Cytotoxicity

In our previous work, we showed elimination of primary acute myeloid leukemia (AML) cells by CD33/CD3 BiTE® antibody construct (AMG 330) mediated cytotoxicity (Krupka et al, Blood 2014). The goal of the present study was to identify innate and adaptive resistance mechanisms to AMG 330 mediated lysis of AML cells. Immune checkpoint molecules have been shown to be a highly relevant escape mechanism of malignant cells to evade innate and adaptive immunity. Previously, it was shown that AML cells upregulate the expression of inhibitory ligands in response to proinflammatory cytokines (Krönig et al, 2014). As AMG 330 mediated T-cell activation induces high levels of the proinflammatory cytokines IFNγ and TNFα, we assessed the constitutive and inducible expression profile of different immune checkpoint molecules on AML cell lines and primary AML cells, including PD-L1, HVEM, ILT3 and SLAMF7 by flow cytometry. No constitutive expression was observed for PD-L1 at time of primary diagnosis in 83.7% of the cases (103/123). In contrast, constitutive expression of HVEM and ILT3 was detected in 73.7% (42/57) and 91.9% (68/74) of patient samples, respectively. Adaptive resistance was evaluated by incubating AML cell lines and primary AML samples with IFNγ and TNFα. We observed an upregulation of PD-L1 and SLAMF7 on AML cell lines and on primary AML patient samples whereas HVEM and ILT3 did not show a significant change in expression level. To test the functional relevance of the immune checkpoint molecules upon AMG 330 mediated lysis, we used an ex vivo long term culture system that enabled us to analyse the dynamic process of receptor-ligand interaction over time. Blockade of the PD-1/PD-L1 interaction resulted in a significantly increase in AMG 330 mediated lysis of primary AML cells (n=9, p=0.03). Currently, blockade of the inducible molecule SLAMF7 in AMG 330 mediated cytotoxicity is being tested. Blocking of HVEM or ILT3 did not result in a significant increase in T cell activation and concomitant lysis of AML cells suggesting a less relevant role of HVEM and ILT3 in resistance to AMG 330 mediated cytotoxicity. The latter might also be influenced by the cytokine microenvironment which favours immune resistance of AML cells.  Using a bead based multiplex assay we screened the bone marrow (BM) plasma from 16 AML patients and 3 healthy donors (HD) for the presence of 33 cytokines. The cytokine profile differed between AML patients and healthy donors (HDs). The plasma levels of IL-8, IP-10 and CXCL-16 were higher in the AML samples compared to those of HDs (p=0.0041, 0.0248 and 0.0289, respectively). In contrast, EGF, FLT3-ligand, RANTES and IL-4 were significantly lower in AML samples compared to HDs (p=0.0227, 0.0145, 0.0041 and 0.0041, respectively). However, we did observe a high inter-patient variability of cytokine composition in AML. To explore the functional relevance of the BM plasma on AMG 330 mediated cytotoxicity, cocultures of AML cell lines and HD T cells were set up using different sources of plasma including fetal calf serum (FCS) and patient derived BM plasma. Interestingly, AMG 330 mediated cytotoxicity was significantly reduced using patient derived BM plasma (n=5) compared to cultures containing FCS (n=4) (mean % lysis FCS 97.4 vs PT 70.6). This was accompanied by a considerable impairment in T-cell proliferation (mean % proliferation FKS 44.7% vs PT 26.6%). Currently, we are investigating which soluble factors are responsible for the immunosuppressive effects and if we can increase lysis efficacy and T-cell proliferation through specific blocking of them. In summary we have identified possible resistance mechanisms of AML cells to AMG 330 mediated cytotoxicity. Dynamic receptor-ligand interactions between target and effector cells as well as soluble factors contribute to AMG 330 mediated lysis of primary AML cells. We hypothesize that AMG 330 mediated cytotoxicity can be augmented through combinatorial approaches including PD-1 blockade. The significance of our findings will first be validated in an in vivo mouse model and prospectively translated into human studies.