Probing the AML Surfaceome for Chimeric Antigen Receptor (CAR) Targets

Adoptive T cell therapy using chimeric antigen receptors (CARs) to redirect the specificity and function of T lymphocytes has demonstrated efficacy in patients with lymphoid malignancies, in particular acute lymphoblastic leukemia (ALL). CD19 CAR therapy can induce durable complete remissions in subjects with CD19⁺ malignancies for whom chemotherapies have led to drug resistance and tumor progression. We previously obtained "breakthrough designation" from the US FDA for CD19 CAR therapy for adult ALL. The success of CD19 CAR therapy bodes well for tackling other hematological malignancies, including Acute Myeloid Leukemia (AML).

The challenge for developing an effective CAR therapy for AML lies in the lack of suitable CAR targets. In ALL, CD19 is expressed on most if not all tumor cells, including tumor-initiating cells, and is only expressed in the normal B cell lineage. In contrast, the AML CAR targets proposed to date (Le-Y, CD123, CD33 and folate receptor-β) do not share this profile. They have not yielded effective and safe tumor eradication in early clinical studies. The clonal heterogeneity of AML and the similarity of AML cancer stem cells to normal hematopoietic stem cells pose additional challenges. Searching for better targets is thus essential. However, the identification of CAR targets is limited by the lack of reliable tools to assess antigen expression across all human tissues.

In order to identify suitable CAR targets for AML, we integrated multiple transcriptomic and proteomic data sources and created an algorithm to identify proteins that can be targeted by single and combinatorial CAR strategies. To comprehensively probe the AML surfaceome, we compiled 474 molecules identified in previous studies and 3,675 molecules we identified by membrane protein biotinylation followed by Mass-Spectometry analysis. To assess whole body antigen expression, we integrated large sets of protein and mRNA expression data and annotated the expression of each AML surface molecule in normal cell types, tissues and organs, including hematopoietic stem cells (HSCs).

We then established a "quality control" filter to assign priority antigens, based on their presence in at least two protein expression datasets (Human Protein Atlas, Human Protein Map and/or Proteomics Database) and their membrane-association. This step yielded 1,694 AML molecules. We further selected molecules with low expression in 64 normal tissues/organs and excluded antigens with high expression in bone marrow HSCs. These steps reduced candidate targets to 215 proteins.

From these, we identified 32 targets overexpressed in diverse AML cells and showing very low overall expression in normal tissues. We performed systematic validation analyses by flow cytometry and identified 11 top candidates with low expression in normal CD34⁺ and CD34⁺CD38^- HSCs and high expression in a panel of AML cells. Further elimination of molecules expressed in T cells reduced the candidate target number to 4, including 2 G-protein coupled receptors not previously reported as CAR targets in AML.

These molecules were however still minimally expressed in some normal tissues, which prompted us to search for pairs of antigens with non-overlapping expression in normal tissues as a strategy to reduce on-target/off-tumor cytotoxicity. We identified 55 such pairs. Our validation studies have so far identified 3 unique pairs of targets showing absent co-expression in normal tissues and ~100% co-expression in a small panel of AML cells. These promising pairs are targeted by T cells co-expressing a CAR specific for one antigen and a chimeric costimulatory receptor (CCR) specific for the other. In a proof-of-principle study, we demonstrate that dual-targeted CAR T cells specific for CD33 and CD70 effectively lyse AML cells with diminished reactivity relative to single-targeted CAR T cells. Further validation studies in larger AML panels are in progress.

This novel discovery approach to CAR target identification should prove very useful to expand CAR therapy applications to AML and other malignancies.