Identification of Novel Targets for Antibody or Chimeric Antigen Receptor (CAR) Based Immunotherapy in AML: Characterizing the Surfaceome from Primary Patient Samples

Background: Despite considerable advances in the field of immunotherapy of acute myeloid leukemia, on-target off-leukemia toxicity remains a challenge. Thus, novel target antigens are sought after. The Cell Surface Capture (CSC) technology enables a discovery-driven approach to detect surface proteins. However, direct processing of primary samples has been limited by the substantial number of viable cells needed for the CSC-workflow. We therefore addressed this issue by improving the technique including the use of our ex vivo culture system. This allowed for an unbiased, direct assessment of primary AML patient samples and thus enabled us to interrogate the AML surfaceome in clinically relevant samples.

Methods: Primary cells from patients with newly diagnosed or relapsed AML were cultured in our ex vivo co-culture system as previously described (Krupka et al. 2014) using MS-5 feeder cells. After 3 days, non-adherent cells were harvested and immediately subjected to the CSC-workflow. Glyco-CSC and its variants Cys-Glyco-CSC and Lys-CSC were initially performed as described by Wollscheid and colleagues (Wollscheid et al. 2009 and Bausch-Fluck et al. 2012) and subsequently modified to improve the yield on AML samples. CSC samples were analyzed by tandem mass spectrometry on an Orbitrap Elite instrument and modified peptides were identified using MaxQuant software.

Results: To enhance the number of successfully identified surface proteins, we adapted the original protocol. These modifications doubled the yield in identified proteins from AML cell lines from initially 125 to 252 in the modified protocol. More importantly, the modifications increased the specificity of the assay significantly. In the original Glyco-CSC experiments, only 54% of all identified peptides displayed a mass shift (of 0.984 Da) associated with successful N-glycosylation and had a transmembrane domain or a signal peptide annotated in UniProt. After modification of the protocol, 80.4% of all peptides fulfilled these criteria. The modified protocol was therefore used for all primary samples. 5 representative primary patient samples from initial diagnosis and 2 samples from relapsed disease were analyzed. All samples yielded sufficient viable cell numbers after ex vivo culture and could successfully be subjected to the CSC workflow. We identified a total of 719 surface proteins fulfilling all filter rules. 22.9% of these proteins had CD annotations. Next, we only considered proteins that were detected by CSC in at least half of the primary patient samples tested. In addition, proteins were filtered to eliminate targets that are abundantly expressed on normal human hematopoietic stem and progenitor cells as well as relevant healthy tissue using publicly available transcriptome databases. 84 proteins were selected as potential candidates for manual screening. Of note, the expression of several antigens currently under investigation for AML immunotherapy (i.e. CD33, CD123, CD135, CLL-1) were detected by the method. We selected 6 promising novel candidate markers previously not described as relevant targets in AML. These were assessed by FACS analysis in independent patient samples. 3/6 of our novel targets showed uniform expression in all independent primary AML samples tested (defined as MFI ratio >1.5).

Conclusion: In conclusion, improvements in the CSC-Workflow combined with our ex vivo culture system allowed for the successful identification of the AML surfaceome from primary patient samples without the necessity of xeno-amplification. We identified 6 novel targets, 3/6 were found to be uniformly expressed in independent primary AML samples. These candidates are now being evaluated further as potential targets for antibody and CAR based immunotherapy in AML.